[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/mlp/ - Pony

[Advertise on 4chan]


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: AltOPp.png (1.54 MB, 2119x1500)
1.54 MB
1.54 MB PNG
Welcome to the Pony Voice Preservation Project!
youtu.be/730zGRwbQuE

The Pony Preservation Project is a collaborative effort by /mlp/ to build and curate pony datasets for as many applications in AI as possible.

Technology has progressed such that a trained neural network can generate convincing voice clips for any person or character using clean audio recordings as a reference. As you can surely imagine, the ability to create audio in the voices of any pony you like has endless applications for pony content creation.

AI is incredibly versatile, basically anything that can be boiled down to a simple dataset can be used for training to create more of it. AI-generated images, fanfics, wAIfu chatbots and even animation are possible, and are being worked on here.

Any anon is free to join, and there are many active tasks that would suit any level of technical expertise. If you’re interested in helping out, take a look at the quick start guide linked below and ask in the thread for any further detail you need.

EQG and G5 are not welcome.

>Quick start guide:
derpy.me/FDnSk
Introduction to the PPP, links to text-to-speech tools, and how (You) can help with active tasks.

>The main Doc:
docs.google.com/document/d/1xe1Clvdg6EFFDtIkkFwT-NPLRDPvkV4G675SUKjxVRU/edit
An in-depth repository of tutorials, resources and archives.

>Active tasks:
Cookie is working on controllable speech
Research into animation AI
Research into pony image generation

>Latest developments:
Public contributions reintroduced for next year's panel (>>37099451)
Singing Talknet models (>>37134971 >>37144858)
Animate automation tool available (>>37147092)
GDrive clone of Master File now available (>>37159549)
New TalkNet models (>>37179832)
SortAnon releases script to run TalkNet on Windows (>>37299594)
A company/group has started selling PPP written scripts for profit.
TalkNet training script (>>37374942)
Fluttershy DeltaVox model (>>37426798)
Delta testing a replacement for guided attention loss for Tacotron2 (>>37453989 >>37458291)
Talk of nVidia RAD-tts (>>37466434)
Young Luna Flash puppet (>>37503317)
Delta updates GPT-J model (>>37554229)
BGM tips added to quick start (>>37601873)
Anonfilly bot update (>>37603355)
Delta GPT-J model (34.147.126.91:3389) (>>37519610)
Latest Synthbot progress report (>>37540665 >>37563548 >>37567441 >>37585347)
Latest Cookie progress report (>>37371181)
Latest Clipper progress report (>>37421445)
Progress on dub
PPP is dubbing S01E01/02
-Doc: derpy.me/8q4Qc
-Spreadsheet: derpy.me/ZCf5i

>The PoneAI drive, an archive for AI pony voice content:
derpy.me/LzRFX

>The /mlp/con live panel shows:
derpy.me/YIFNt

>Clipper’s Master Files, the central location for MLP voice data:
mega.nz/#F!L952DI4Q!nibaVrvxbwgCgXMlPHVnVw
mega.nz/folder/0UhSmYAB#WBrB-qCprQTofkAhwMp5CQ

>Cool, where is the discord/forum/whatever unifying place for this project?
You're looking at it.

Last Thread:
>>37596051
>>
FAQs:
If your question isn’t listed here, take a look in the quick start guide and main doc to see if it’s already answered there. Use the tabs on the left for easy navigation.
Quick: derpy.me/FDnSk
Main: derpy.me/lN6li

>Where can I find the AI text-to-speech tools and how do I use them?
A list of TTS tools: derpy.me/A8Us4
How to get the best out of them: derpy.me/eA8Wo
More detailed explanations are in the main doc: derpy.me/lN6li

>Where can I find content made with the voice AI?
In the PoneAI drive: derpy.me/LzRFX

>I want to know more about the PPP, but I can’t be arsed to read the doc.
See the live PPP panel shows presented on /mlp/con for a more condensed overview.
derpy.me/pVeU0
derpy.me/Jwj8a

>How can I help with the PPP?
Build datasets, train AIs, and use the AI to make more pony content. Take a look at the quick start guide for current active tasks, or start your own in the thread if you have an idea. There’s always more data to collect and more AIs to train.

>Did you know that such and such voiced this other thing that could be used for voice data?
It is best to keep to official audio only unless there is very little of it available. If you know of a good source of audio for characters with few (or just fewer) lines, please post it in the thread. 5.1 is generally required unless you have a source already clean of background noise. Preferably post a sample or link. The easier you make it, the more likely it will be done.

>What about fan-imitations of official voices?
No.

>Will you guys be doing a [insert language here] version of the AI?
Probably not, but you're welcome to. You can however get most of the way there by using phonetic transcriptions of other languages as input for the AI.

>What about [insert OC here]'s voice?
It is often quite difficult to find good quality audio data for OCs. If you happen to know any, post them in the thread and we’ll take a look.

>I have an idea!
Great. Post it in the thread and we'll discuss it.

>Do you have a Code of Conduct?
Of course: 15.ai/code

>Is this project open source? Who is in charge of this?
derpy.me/CQ3Ca
>>
File: ASmallAnchorUUUU.jpg (35 KB, 284x300)
35 KB
35 KB JPG
>>37634877
Anchor.
>>
File: AIdub.png (156 KB, 896x504)
156 KB
156 KB PNG
PPP FiM AI REDUB
We're redubbing the first two episodes of FiM with AI, and we want the whole board to take part. See the Info Doc, and this youtu.be/X2uPYztiR_g for more details.

The Info Doc:
https://docs.google.com/document/d/1UItM5uzhzVXfq07YqIWcHHJ2HnfbqNBj-xNJo5Xu_hY
The spreadsheet:
https://docs.google.com/spreadsheets/d/1M9CJ3SppL44yHziC4DWzkS0s3qOeEtzT85fO3qjdnlc

THE MODELS:

DeltaVox:
https://drive.google.com/drive/folders/1xGpP_wYxMCO-ZtKA5aNDXGnRTOZLwEWe
DeltaVox extra characters: (extract to DeltaVox RS/models)
https://drive.google.com/drive/folders/1SuQPDmAuuF1pmemH2uzQCBdhnFPaMJe7

TalkNet Online:
https://colab.research.google.com/drive/1aj6Jk8cpRw7SsN3JSYCv57CrR6s0gYPB
TalkNet Offline Installer (REQUIRES NVIDIA 4GB GPU/900 SERIES OR NEWER):
https://github.com/SortAnon/ControllableTalkNet/releases/latest/download/TalkNetOffline.zip

Current 15.ai link:
https://finalfinaltest.15.ai/

We're getting close to the end, so we might want to think of a contingency plan if people run out of time on the scenes. Please try to deliver as soon as possible if you're still working on a clip.
>>
>>37634888
I will hopefully have the Spike sleeping scene done tomorrow.
>>
File: 1630222906348.png (565 KB, 600x508)
565 KB
565 KB PNG
>Thread 99
https://youtu.be/jxvR7ZUjaJk
>>
No, but seriously, thread 99.5, I mean it just makes sense.
>>
>>37634596
>>37632871
>>37634506
Ok, it's still not perfect, but I managed to improve the coodination to the background music!
I fiddled with the tempo and the version of the background music, so I hope you guys like the updated version. :)
If there is anything that you think I can realistically improve, please tell me!
(If you'd like the audacity files, I can post those too)

https://u.smutty.horse/mdluqfasosr.mp3
>>
File: Capture.png (24 KB, 1804x612)
24 KB
24 KB PNG
>>37634877
Has anyone else noticed audio playback on 15.ai is weirdly quiet? Pic related, top waveform is audio captured from the site, bottom one is the downloaded file.
It's annoying to have to keep raising and lowering my volume when I switch from generating lines to editing them together.
>>37635249
That's fucking gay. We're at 82%, and have 493 (-1) posts before this thread reaches bump limit. We have more than enough time to finish.
>>
File: coco gets screamed at.png (577 KB, 1025x650)
577 KB
577 KB PNG
>99
>>
File: Raisin Brain.png (851 KB, 4605x3000)
851 KB
851 KB PNG
Maybe You Need... (Preview) ------> https://u.smutty.horse/mdlvqfjxtji.mp3 (1:41)

Aren't you forgetting something...?

----------------

First part of a WIP audio focusing on Gilda. This is the only part of the audio that is more or less completely finished, so I decided to post it to show it's being worked on. Second act is still in the works, and third act hasn't been written or worked on yet. I'm gonna try to get this all done in time for Thread 100, or 101 at the latest.

Gilda's a tricky character to get down properly, but I think she deserves some love.
>>
File: report AND IGNORE.png (403 KB, 1501x929)
403 KB
403 KB PNG
>>37634883
PSA reminded to not feed the tranny trolls.
>>
>spend all day generating AI audio
>now I can't imagine their normal voices without that "AI" audio ripple effect
help
>>
>>37635730
watch the show
>>
>>37634888
Volume normalization will be handled, right? Some clips are obviously louder or quieter than others.
>>
Thanks to some anons from the last thread, I have ideas as to how to finish my final version of 1-27. But I've been tasked with moving furniture and boxes all day today and am exhausted. So I'll probably start either tonight or tomorrow.

>>37635414
We just need a volume slider on the site, finally. So I don't have to have to turn my entire system sound down to 10% so ponies aren't screaming at me.
>>
>>37634888
1-1 Series intro
1-7 Elements of harmony
1-14 Apple brunch
1-16 Wonderbolts
1-24 Revalation
2-4 Sleeping Spike
2-5 Twi gets confronted
2-10 Spooky trees
2-16 NMM "wins"

Anons working on those clip, as you can see we are currently on thread 99, I do not want to stress you guys out with having a roughly seven day to get to finish so lets talk about this from different angle, what do we fellow anons can do to help you guys with the creative process of getting those clips done in the timely manner ?
>>
>>37635817
A firm kick in the nuts so they stop slacking off.
>>
>>37634888
3 out of 9 contingency clips completed. For the sake of not confusing things I'll only upload them if/when needed.

>>37635817
>what do we fellow anons can do to help
Seconding this. If there's anything I can do to aid in the finishing of the remaining scenes, especially in the audio/music side of things, I'm happy to help.
>>
>>37635940
>For the sake of not confusing things I'll only upload them if/when needed.
Will you post the unused ones after the project is done, as bonus outtakes?
>>
>>37635635
Fucking birb.
>>
>>37635635
This is really fucking good. I can't wait to hear where this goes.
>>
has anyone tried making a talknet of this video?
https://www.youtube.com/watch?v=IHQH57y2d7Y
>>
>>37634897
I think *BGM* knows exactly how much time he needs to edit something.
>>
>>37636237
We don't have any singing models of male voices, and anyway I liked the original Muppet Movie version better.
https://www.youtube.com/watch?v=Q6NQcO9KTBY
>>
>>37636264
your taste's pretty good, but i meant only for twilight since the other vocalist sounds a surprising amount like mordecai already. unless that's not how it works and you'd have to replace all vocals?
>>
>>37636271
No, it's possible to replace only Twilight's voice. It all depends on how well the vocals can be isolated from the music.
>>
>>37636279
oh cool! i'll leave it to people who know how, if they're into the idea.
>>
>>37635817
I say we open up all the ones that are "being worked on" so that we can get something as a fallback. Then if the original anons get theirs in in time we can take those instead.
>>
>>37636304
First I would like to see if within next day or two any of those anons respond if they keep working on those in the thread, while im sure BGM will do a superb work on the backups I would still want to see what ideas the original anons had when they picked their clips. BUT on the other hand I do kind of want to make my own version of few of those clips as well
>>
>>37636469
nothing is stopping you from already making backups if you have an idea floating around.
>>
File: RiM.jpg (384 KB, 2000x1200)
384 KB
384 KB JPG
>>37635817
>>37635940
1-16 Anon here, could you please help me with the music side of things? I want to change the melody of the guitar riffs to the one of the riffs in this modern classic:
https://u.smutty.horse/mcbvbohuldg.wav
Trouble is, I have no idea how to do this, and I'm not optimistic about my chances of mastering music production in less than a week.
For reference, here are links to the original riffs in Clipper's Master File (they have some strange distortion on the percussion, but the guitar is pretty much intact):
https://mega.nz/folder/0UhSmYAB#WBrB-qCprQTofkAhwMp5CQ/file/QRpA1ajL
https://mega.nz/folder/0UhSmYAB#WBrB-qCprQTofkAhwMp5CQ/file/NRwWzIRJ

Any help would be appreciated, though I understand if this is too much to ask on such a short notice.
>>
>>37636237
>>37636264
When teaching your pone to sing based off of male voices: Remember to modify the vocals track or increase the pitch within talknet, or your ponies will sound weird.

>>37636279
I'd really like to learn how to separate vocals as well. I would love to get AJ and Rarity to sing "Islands in the Stream", but don't want to touch that until I can separate Kenny Rogers from Dolly Parton.
>>
>>37636505
that song was originally posted in tempo thread 37210167 by anon 37254350 Wed 14 Jul 2021 , I can ask in the thread if the original song creator could share the instrumental for this project
>>
Whoever made the rarity/coco and fluttershy/pinkie hoofball tracks you're fantastic and I look forward for your next work.
>>
>>37636557
That would be nice, but I was thinking of recreating the riffs with a guitar that sounded closer to the one used on the show. You know, change the melody but keep the same instruments.
>>
>>37636578
ahh, sorry bud, I've misunderstood what you meant in the first post. yeah i dont think there is amount of music/video editing that can be applied in here, a proper musicfag would need to make that from scratch
>>
>>37636578
For what it's worth, here's the instrumental:
https://u.smutty.horse/mdlzmkbczgk.wav
>That would be nice, but I was thinking of recreating the riffs with a guitar that sounded closer to the one used on the show. You know, change the melody but keep the same instruments.
I can re-record or change the amp sim. Could you be more specific about what you want?
>>
File: Spoiler Image (5 KB, 435x412)
5 KB
5 KB PNG
>>37636705
Thank you!
>I can re-record or change the amp sim. Could you be more specific about what you want?
I'll level with you, I have no clue when it comes to music. If I had to describe it, I'd say I want something more peppy and less base-y, like the show itself.
>Captcha: STR8S
IT KNOWS
>>
>>37636758
Like, could you give files/timestamps for reference?
>>
File: P0VYD.png (6 KB, 299x121)
6 KB
6 KB PNG
MARESMARESMARESMARESMARESMARESMARESMARESMARESMARESMARESMARESMARESMARES
Captcha: P0VYD
>>
>>37636775
Sure, all I want is to have a guitar that sounds like the one in this clip, at 0:25 and 0:52:
https://u.smutty.horse/mcstpriwfxv.mp4
Use that guitar to play the same riff as the one from the file you posted, between 0:04 to 0:08.
>>
>Delta GPT-J model (34.147.126.91:3389)
Can we teach it to discuss the show?
>>
>>37637029
I think it would be more fun to teach it to script an entire episode. Then we can generate all the lines and animate it purposefully poorly then brand it as a "deleted/lost episode"
>>
>>37637059
No, a quality show discussion thread would be much better in spirit with our fandom being tiny but generating 500x times more content than other fandoms 5000x times bigger than us.
>>
>>37636566
Could you link those for an anon with blue balls?
>>
>>37637099
https://u.smutty.horse/mdipqxaiaui.mp3
https://u.smutty.horse/mdkbbjzvknm.mp3
>>
>>37635635
Can't wait, that bird squawk might need a little editing though, took me out of it.
>>
>>37637111
Thanks man
>>
>>37635817
>what do we fellow anons can do to help you guys with the creative process of getting those clips done in the timely manner ?
2-16 here
If someone to whom TalkNet isn't as much of a bitch (or who simply is awake when I'm not) can get me Pinkie singing the chorus to Somebody That I Used to Know (just the vocals and just the chorus specifically, and you don't need to stitch them up or time them right to the song together, a few different files are fine for what I need it for) that would help.
A better PC would also help but that's not exactly something you can help with.
Also, aside from the one local schizo ITT, how do the people this will be shown to feel about G5?
>>
>>37637134
I was going for a bit of a cartoony feel with a couple parts, especially since Gilda does a couple squawks here and there; it's mainly there to show she's reacting in a "woah" to whatever the viewer said, but I'll see if I can make it a bit more immersive without taking listeners out too much.
>>
>>37637534
>spoiler
Are you planning to include something relating to G5 in your clip? Or is this just an question unrelated to 2-16?
I believe you can make anything funny if you tried it clever-y enough, but other than an offhoof comment about it, I'm not sure it'll go over well. Personally, clips or pictures of G5 is a no-go. This is FIM.
But, If you're just asking about our opinions about G5, then it's OK.
>>
>>37637534
>spoiler
Very strongly against it. There's time and place for everything and redub of the first two episodes is no place for G5.
>>
>>37637534
>spoiler bit
no, there is no reason to add it here for any reason (whenever you want to make a fun of it or shit on it or whatever).
>song bit
Just to be sure, do you want the vocal bits separated from instrumental right? (I may may try to get you those lines song tomorrow, unless someone else beats me to it in the meantime).
>>
>>37637653
>Just to be sure, do you want the vocal bits separated from instrumental right?
Correct.
>G5
I'll make two different versions, one for the episode and one for personal enjoyment. I'll post both here.
>>
>>37637567
If it was more Gilda like rather than royalty-free-parrot-sfx I think it would have been fine.
>>
>>37637534
>spoiler
Does not belong at all, especially if you are talking about a clip submission.
>>
Here is 1-14, Apple "Brunch":
https://files.catbox.moe/gpwlmn.mp4
I'm using Catbox because Smutty refuses to collaborate.
If there's anything you notice that should be improved or fixed or changed, tell me and I'll see what I can do. Hopefully it's enjoyed, and sorry for the late delivery.
>>
>>37635635
that knock scared the shit outta me
>>
>>37635635
This is TalkNet I assume
>>
>>37637979
If a voice sounds emotive, it's usually 15.
>>
>>37638070
I thought we were supposed to credit the correct TTS in our projects
>>
Hello! I've been using the Controllable Talknet and I've run into a problem I would like to know more information about.
I am using Twilight Sparkle's singing voice for this example, but I intend on using Celestia's voice for the project I am doing. I will include her version too, as they're both seemingly bugged as well, and knowing how to fix both could help in future endeavors.
I am trying to make them say "It's always such a pleasure" with my Reference Audio. I would assume this would of turned up fine, but it didn't output the way I wanted;

Original Reference Audio: https://u.smutty.horse/mdmcaorqtpe.wav
Twilight Sparkle (Singing): https://u.smutty.horse/mdmcasadxkg.wav
Celestia (Talking): https://u.smutty.horse/mdmcatiblvq.wav
ARPABET: {IH T S} {AO L W EY Z} {S AH CH} {AH} {P L EH ZH ER}

This might be an easy fix but I am very new at this sort of thing. I have a few ideas myself, but I would love to get some help from the main thread first. Thank you.
>>
>>37634888
I ended up cutting back on my plans a bit, but here's 2-4 sleeping Spike.
https://u.smutty.horse/mdmcgclgppr.mp4
>>
>>37637817
make sure you and other link their post to the main organization post >>37634888 so docanon can check it out on the list.
>>
>>37636827
I'm still not really sure what your intent was (the predominant sound in the video was a lead guitar with a very faint rhythm guitar, while my riff is very rhythm guitar centric). I scooped out some of the rhythm guitar, made the bass a bit brighter, and added a lead that comes in later (the solo also has the same sound).
https://u.smutty.horse/mdmcjoerxto.wav
>>
>>37638091
For Twilight singing, you could try to write the ARPAbet as {P L EH ZH ZH ZH ER}. The problem with Celestia may be that the model can't deal with the harmonics in GlaDOS's voice, and trying to use a speaking model for singing usually doesn't produce good results.
>>
>>37635940
4/8 contingencies completed.

>>37636003
I see no reason not to.

>>37636469
>>37636496
Also this. If anyone wants to contribute to making backups feel free. We already have them in place for 1-1, 1-7, 1-17, and 1-24. We still need them for any clips not delivered, those currently being 1-16, 2-5, 2-10, and 2-16. I only ask that you preferably email them to me so I have them in one place heyitsmenicholascage@gmail.com but otherwise just keep the exact links off thread until they're necessary, for the sake of not confusing things. I'd be happy to throw together a 'deleted scenes' compilation with any that don't get used.
>>
File: 1630331728905.gif (1.09 MB, 640x360)
1.09 MB
1.09 MB GIF
1-7 Elements of Harmony here, I'll deliver it later today, worst case tomorrow. Draft is done, just seeing if I can add some juice to it.
>>
>>37635817
2-5 here if you want to go to work for me that'd be great. Audio is done, so worst case scenario I can just deliver that if we get too close to the end. I wanted to use premiere for a few clip edits but I suck balls.

I'll deliver before Sunday or 400 posts or I'll forfeit it.
>>
File: 256470.gif (2.23 MB, 542x598)
2.23 MB
2.23 MB GIF
>>37637817
I highly appreciate the continuity from the previous clip (which happens to be mine, lol), and this one is just perfect because where mine was kinda easing into the idea of slavery subtly, yours abruptly slammed into the idea and exploded!

Subtlety meets sudden bluntness is great comedy.

AJ's voice had decent quality overall (could still be cleaner with the right evolution and generation to match my results, but not that important). That sped up part sounded pretty strange though. Maybe use an anxious or high energy Deepmoji context to get a faster raw result? But yeah, awesome work anon!
>>
Now that I'm finally done with 3 days of constantly moving furniture, I can finally get started on the final version for 1-27. Had an idea to change the words of the credits, it was rough finding a similar enough font for free considering the original is apparently almost $140, which is ridiculous. But I guess there's a reason they chose it. Still brainstorming exactly what I wanna change the text to though.

>>37637817
Man, I'm still dying over "Actually, the preferred term would be filthy mu-" and "Will you not consume my flesh?" *sad face and music*
>>
text completer is offline again
might as well repost this
>>
>>37637979
Sorry, was away from my computer for a bit. This is all 15ai, as Gilda hasn't been added to the TalkNet yet, at least at the time I last used it a few weeks ago.
>>37638079
15 has stated that if you use his site for any projects, you can only use his software for that specific project as to not confuse listeners between various TTS sites and software that may sound similar, as it may harm the overall goal he's striving for. For people who are close to the PPP, I think there's exceptions (TalkNet), as long as you specify it.
>>
>>37637972
You feel immersed yet, Anon?
>>37637789
I was actually trying to look for squawk noises that would sound like they came from Gilda, but hadn't been successful. The audio file I downloaded has a few other squawks that I haven't heard yet, so the one here will probably be replaced as soon as I find a more appropriate one.
>>
>>37639488
>>37639496
Good work friend.
>>
>>37634888
>>37630121
Finished 2-10 Spooky trees:
https://u.smutty.horse/mdmftypkcwh.mp4

Are we going to do anything special with the Ad Breaks?, I was thinking of something similar to the "Pinkie Pie Says" clip.. I would show it but I can't seem to find it right now.
>>
>>37639723
>Are we going to do anything special with the Ad Breaks
We're behind already on what clips we have, people should try and focus on those. As for the ad breaks themselves personally I think we should just keep to the episode.
>>
>>37639723
>Are we going to do anything special with the Ad Breaks?
I've thought about it, though as >>37639770 said adding more to it with an already fast-approaching deadline is probably not a good idea.

>>37638614
>if you want to go to work for me that'd be great.
What are you looking to do with your clip, and what exactly do you need help with?
>>
>>37639488
I'm pretty sure when they said "no mixing", they meant like, no using 15ai results to use as training data for other AI? As in no mixing 15 with another. But that's just my interpretation.
>>
>>37640171
No, the rule was specifically added to prevent people from mixing shit TTS like vocodes and uberduck with 15.ai voices.
>>
>>37639488
>>37640177
Correct, that rule was made to stop twittermites from using 15.ai with shitty vocodes Spongebob and slapping "15AI" on the video as it makes rounds on the internet. I'm pretty sure 15 said that it's fine when >we do it for use in the PPP thread content.
>>
File: 1626666851694.gif (4 KB, 90x51)
4 KB
4 KB GIF
>>37634888
1-7 Elements of Harmony is done.
https://u.smutty.horse/mdmhpeyzqez.mp4

Not sure if the explosion gag at the end is too intrusive when transitioning to the following clip, if it is I can upload a version without it.
Also, fuck keyframe masking.
>>
> Featuwed Voice Perfowmows

>Tawa Stwong as Twiwight Spowkew
>Ashwie Bow as Appewjack and Wainbow Dash
>Andwea Wibman as Pinkie Pie and Fwuddershy
>Tabitha St Gewmain as Wawidy
>Caffy Wesewuck as Spike
>Nicow Owiver as Pwincess Cewesdia
>Mashew Cweeber as Apple Bwoom
>Madewin Petews as ScootAWOO
>Cwaire Corwett as Sweedie Bew

The edits for the final version of 1-27 have begun...
>>
File: 540237.png (1012 KB, 1125x698)
1012 KB
1012 KB PNG
>>37636237
>>37636271
I made an instrumental track, if anyone wants to try getting Twilight to sing over this.
https://u.smutty.horse/mdmihchjeen.wav

The separated vocals are way too noisy to be used as reference audio, but here they are anyway.
https://u.smutty.horse/mdmikxzfhgg.wav

I'm tone deaf and only have a laptop mic, but I'll try to record myself singing it later and post results
>>
>>37640457
That reminds me: There's a new state-of-the-art for separating vocals.
https://github.com/bytedance/music_source_separation

Vocals: https://u.smutty.horse/mdmjfxsdsro.ogg
Accompaniment: https://u.smutty.horse/mdmjfxszwea.ogg
>>
>>37640626
Holy Shit, this is impressive the instruments are almost perfect... Vocals seem better than SpleeterGUI
>>
File: 1626821667850.png (122 KB, 729x600)
122 KB
122 KB PNG
>>37637534
>>37637676
https://u.smutty.horse/mdmjncckjui.zip
Here, the Pinkie singing the chore bit plus the original studio vocals clips (in case you or someone else wanted to regenerate those). There was a minor problem with the word "that " in the 7th clips, for some reason the talknet wasn't detecting it at all so I've just remove it from the text generator, hopefully it wouldn't be that noticeable in final result.
Also I would suggest to do some minor audio wizardly to make it sound better.
>>
>>37640457
>>37640626
Here's the "Movin' Right Along" audio with Twilight and Spike. I could try to edit it back into the video later.
https://u.smutty.horse/mdmkdcnuqna.wav
>>
This too way longer than expected to script up. Here's the text showing the 1-27 credit words replacement.
Feel free to proofread it if you want: https://u.smutty.horse/mdmkieorfkw.txt (Warning: uwu-speak translation. High cringe levels.)

I'll get started on the actual edit some time tomorrow
>>
>>37640962
I'm sure no one would notice a mistake.
>>
>>37626945
Sunny training run failed with bad results after I manually took out noisy clips myself. I tried using ASR + Levenshtein against ground truth transcripts as a way to automatically identify noisy clips, exploiting the fact that STT struggles on noisier audio, but that didn't work. VITS shows really good potential on multispeaker so I'll likely train a big one on all MLP* speakers, but will have to plagiarize Cookie's ngroks as way of offering a front end for inferencing because no export. It also sounds less flat and more emotional than (stock) Tacotron 2 to me.
https://jaywalnut310.github.io/vits-demo/index.html
>>37639346
Looks like instance got a restart or something, starting it now.
>>
>>37641810
Yeah, one movie's worth isn't enough data. Could try sending it over to 15 to see what he can do with it in his system? Honestly though, we need a good season or two at the very least to get something viable.
>>
>>37642109
I say we wait for VAs to get confirmed first. I don't think it's gonna be a good use of effort to start the data distillation process this early on if we have to toss even some work out the window if certain VAs don't renew.
>>
>>37635697
anons, we know its shit, they know its shit, sadly it wouldn't change their invasive behavior in here.
Anyway, anons doing working on clips, you got everything you need to finish your bits?
>>
>>37639723
Good job, but the pan noise is pretty loud... and the ending of the clip gives me flashbacks to the anti-anthology. Pictures on a black background doesn't fit well here-- it sorta breaks the immersion.
I know this project is starting to run into a tight schedule, but if the faces could instead be masked onto the trees, that'd make it 10x better. (And if you can't, I could probably mask it for you if you need the help. Shouldn't be too long to do.)
>>
2-16 status update. I'm done on the audio side of things, now I only to add a bunch of pictures to one scene and then hope my PC doesn't kill itself when I render the video. Maybe I'll do some fancier editing on that section too, but probably not. Not really sure about doing a separate version for personal enjoyment either, maybe at a later date when my wrists won't be hurting from moving dozens of files around.
>>37640699
Thank you for this, by the way. I didn't thank you sooner only to avoid wasting more thread space.
>>
>>37641810
Address is still http://34.147.126.91:3389 right? I can't access it.
>>
>>37635635
Nice, sounds almost natural.
Love me some Gilda.
>>
>>37642302
Is this better?
https://u.smutty.horse/mdmohzxjmon.mp4
>>
File: Movin' Right Along.jpg (165 KB, 1280x720)
165 KB
165 KB JPG
>>37634883
>>37636237
>>37640889
And here's the dubbed "Movin' Right Along" video. Seeing the animation really does help sell the voices.
https://u.smutty.horse/mdmpdcushkd.mp4
>>
>>37639247
>Had an idea to change the words of the credits, it was rough finding a similar enough font for free considering the original is apparently almost $140, which is ridiculous.
The font used in the show's credits is Woodrow W00, which you can get here:
https://www.onlinewebfonts.com/download/7a254c051bfbb6a604d602768b0d1cec
>>
File: 1619635719266.jpg (191 KB, 1600x1538)
191 KB
191 KB JPG
>>37643089
Yeah Twilight's voice, even if it's a bit on the scratchy side, still makes the animation really pop a lot better. Also makes the original voice sound a lot more like Mordecai.
>>
>>37643089
>https://u.smutty.horse/mdmpdcushkd.mp4
pretty good tho i think it could sound better
>>
>>37640171
>>37640177
>>37640206
Im a lurker but I think in the end credits or something you guys should specify which scenes were made with which TTS so there's no mixup and proper credit is given. If you have that info that is.
>>
>tried to get Fluttershy to do OG Loc's rap
>ended up sounding sexual
>>
>>37643580
You can't just say that and not upload it.
>>
>>37643623
I was just fucking around for a while, I didn't save anything or even finish the full song.
>>
>>37643641
Then why'd you even post in the first place? You know damn well that people appreciate that kind of content by now. Fucking useless idiot.
>>
>>37644012
ngl was hoping for a response like yours actually :^)
>>
i see it still struggles with snails
https://files.catbox.moe/267gd3.wav
>>
how are those new characters going 5*3?
>>
>>37641810
I still can't access the site
>>
File: flutterpaint.png (12 KB, 100x100)
12 KB
12 KB PNG
>1-27 credits text edit is now in the rendering phase
>Added extra jokes for extra entertainment value "BUY OUR TOYS"
>Combine with original edit without shadows as emergency final version in case full isn't out by deadline
Things are coming along smoothly now
>>
>>37634888
>>37644807
Updating 1-27 - https://u.smutty.horse/mdmvtgrxkjo.mp4
If there are no further updates/tweaks/whatever from me, consider this the final version.
>>
infinity in mod infinity is 0
>>
File: majestic.jpg (1.58 MB, 2800x1800)
1.58 MB
1.58 MB JPG
>>37634888
Hey, I went through the spreadsheet looking for segments with video unaltered from the original,
and after finding them, muxed their audio with the original video tracks.
This fixes clips which have an invalid resolution and/or framerate, and will also lessen generation loss on the video side of things.
The clips are, in order of appearance:

1-13
u.smutty.horse/mdmwbmevwmt.mp4

1-14
u.smutty.horse/mdmwbonbzfq.mp4

1-26
u.smutty.horse/mdmwbpgwkeh.mp4

2-3
u.smutty.horse/mdmwbnmlvdf.mp4

2-14
u.smutty.horse/mdmwbowpolx.mp4
>>
>>37645102
I have already taken care of this. Thanks for the effort though.
>>
>>37645102
Which rendering option *are you* actually supposed to use to get the right framerate (for someone using Vegas specifically)?
>>
>>37645377
1080p/24Hz.
>>
>>37644949
Why must you do this? Can't we just use the old version. This is absolutely disgusting.
>>
>>37644949
just beautiful.
>>
>>37644949
The old version was so much better.
>>
>>37644949
No, absolutely not.
>>
>>37645852
Well, the technology is still an infant. I think in a few years, 15ai will be leagues better than anything we've seen since 2019. As for the 'abridged episode humor', I'm inclined to agree, but mainly because we're trying to abridge an existing episode. I feel that if a bunch of Anons got together and made a new episode from scratch, it'd be better written and paced.
>>
>>37645852
Are you confusing Talknet with 15.ai? Because the robotic sounding voices are from Talknet and it's definitely anons' fault that they're mixing the voices, which 15.ai specifically says not to do on the site.
>>
>>37645852
Honestly I agree with the voices sounding like robots, BUT I think it’s because they were made using TalkNet. I was on the /mlp/ TF2 server last night and one of the AJ TalkNet songs was being played over the mic and they honestly sounded terrible but only a couple people seemed to notice.
I can almost immediately tell that something was made using 15 compared to something that was made using TalkNet but it seems that most people can’t tell the difference between AI songs and real songs for some reason.
>>
>>37645852
Hard agree. 99% of AI songs sound like shit and robotic as fuck. Don't know why everyone loves them so much. I'm not even a fucking musician but even I can tell that the singing is completely off tune.
>>
>>37645978
Are you telling me that you seriously can't hear the glitchy artifacts in the AI songs? You have to be literally deaf to not be able to hear it. Absolutely wild.
>>
>>37645377
Don't listen to the faggot who doesn't know what he's talking about. After you've selected the right MP4 format, click on customize template and pick the correct framerate.
>>37645852
Correct, but anyone on this board who honestly expects people here to produce high quality humour is a genuine retard and the tools being comparable to splicing is still a great thing.
>>37644949
Absolutely perfect and the seething ITT is extremely sweet.
>>
>>37645881
Agree, TalkNet is cool and all for small projects but if you’re going for something relatively serious like an episode redubbing, it feels so jarring to hear TalkNet Twilight right next to 15 Twilight, it throws any immersion right out the window.
>>
>>37645993
>the seething ITT is extremely sweet.
>from the thread it's supposed to represent
That should be reason enough to reject it. The previous submission for the segment was far better anyway.
>>
>>37645852
>>37645906

I mean, the reason we're all excited is because a lot of us were here for the Fall 2019 breakthroughs. TwAIlight basically sounded like a robot, and hardly sounded like Twilight at all, but it was exciting because we could *kinda* hear her through it, and that was a big deal! We're excited now because, yeah, the singing sounds a bit robotic still, but it's showing promise. And the regular vocals out of 15 are absolutely stellar. We've come a really long way and there's still a bit of a ways to go, but it's remarkable to have come this far.
>>
>>37645906
>>37645931
>>37645988
At first I thought people were just giving a pass because it’s a new cool thing that popped up but some of them are actually seriously trying to argue that the AI songs sound better than humans singing. Like wtf?
And they’re being 100% genuine too. They seriously listen to an AI song and think that it was sung by a human no matter how glitchy or off tune it sounds.
>>
>>37645852
Could you elaborate?
>>
>>37646003
>PPP is one person
>>
>>37646176
I saw three other posts besides mine at least. Even more if you count the previous time that thing was posted.
>>
>>37642707
Yes! Good job :)
>>
>>37642513
>>37644187
Hmm. New IP: http://35.204.47.23:3389/
Also my TPU trial is going to run out in ten days or so so Colab better fix its shit or it's going to be no GPT-J anymore.
>>
>>37646239
any chance at having that downloadable model posted here (I think someone mentioned few threads back a pytorch conversion that works on weaker gpus)?
>>
>>37646251
Gimme a sec.
>>
>>37645906
When we first started the redubbing project, 15 hadn't put his site back online yet, so everyone had to use TalkNet. And we knew at the outset that this was going to end up sounding like an abridged series. Dubbing songs with TalkNet is mainly good for the novelty of hearing the Mane 6 singing popular songs, and we know they sound robotic, but they're the best option we've got because 15'S MODELS CAN'T SING. It might be interesting to see if 15 can apply his method to voice style transfer, but he could be sidetracked for years trying to work on that. The singing models probably sound robotic because of the small size of the singing datasets compared to the speaking datasets. The only way we could improve them would be seeing if there's a multispeaker implementation of TalkNet.
>>
>>37646251
>>37646292
Here are the slimmed down weights in JAX, 11 GB .tar
https://storage.googleapis.com/eudsk/tfmlong.tar
>>37644949
uh oh
>>
Basically until you fix this >>37635730 your voice generation program is no better than Google's.

>>37646058
While it sounds slightly better than the tiny YTP audio spliced phrases, it's really not that much of an improvement to call it superior technology. We've had audio splicing since forever. Only worthwhile thing about it is that it can generate new words instantly for a few specific voices, but it's not that much better than Google's text to speech even with the added emotion. And when the AI really doesn't want to say something it really goes off the rails with the pronunciation and ruins any tiny amount of smoothness it had, which happens a lot.
https://www.youtube.com/watch?v=nR16_FNosms
https://www.youtube.com/watch?v=ad8XLwuYxpk
https://www.youtube.com/watch?v=wO1JNRc6A5g
https://www.youtube.com/watch?v=CutNWwYt4Mg

We've also had voice synth modification since the dawn of time.
It's sad how some people in this thread say the songs produced by 15.AI sound good when the 15.AI exclusively uses talking samples not meant for singing.
https://www.youtube.com/watch?v=viTkj0eu-fk
https://www.youtube.com/watch?v=ES7gaZA4b8U
https://www.youtube.com/watch?v=eG4DByc3d8A

The non-English text to speech might fool EFLs into thinking it sounds smoother than their native English language.
https://www.youtube.com/watch?v=5WOKoFMRr5Q

We can't call it: an evolution, the future, a great stepping stone in voice generation. It's just a sidegrade.
>>
>>37646340
>It's sad how some people in this thread say the songs produced by 15.AI sound good when the 15.AI exclusively uses talking samples not meant for singing.
Oh so you're basically admitting that you were confusing TalkNet with 15.ai
God damn you are a fucking retard
>>
>>37646340
You moron, songs are produced by TalkNet
>>
>>37646340
>but it's not that much better than Google's text to speech even with the added emotion
The emotion IS the HARDEST PART, brainlet. No one has been able to nail the emotion like 15, not even Google.
>>
File: 1601630748456.jpg (100 KB, 1000x1000)
100 KB
100 KB JPG
>>37646340
Retard.
>>
File: 1620455476139.png (719 KB, 1001x932)
719 KB
719 KB PNG
>>37646340
Thanks for outing yourself as a newfag, newfag
>>
>>37646340
While obviously this guy is a retard, at least it got people talking about the TalkNet models, which I thought sounded pretty bad too. I thought I was the only one who preferred the 15ai songs like Pony Zone over the TalkNet song covers, so it’s good to see that others think the same.
>>
The gpt j inference is still unreachable for me
>>
https://u.smutty.horse/mdnaeqkutxj.wav
>>
File: GlimPosting.png (162 KB, 754x1024)
162 KB
162 KB PNG
https://u.smutty.horse/mdnagdbctvn.wav
>>
>>37646239
Because I have no idea how to correctly prompt this, thought I'd just give it a go. Separated lines by "/" to make it a little neater

>Write a complete script of a new episode of my little pony friendship is magic.

>"Once upon a time, in the magical land of Equestria"


Author's Note
I'm sorry to disappoint you if you don't like this, but I hope you enjoy my attempt.
I'll give a prize to anyone who gives me a good name for the episode.

<|endoftext|>1. Curiosity
2. The Siege
3. The Calm Before the Storm
4. The First Attack
5. The First Casualties
6. The Assault
7. The Chase
8. The Champion
9. The Will of the Gods

<|endoftext|>Who is the Spectral Mare? / “I’m sorry, but I must ask you to leave.” / “Excuse me?” / “I said, you must leave.” / “Who do you think you are?” / “I am a former friend, one of your worst enemies.” / “Why should I listen to you?” / “I am the one who has been ordered to remove you from the area.” / “And why would you do that?” / “The rules of the game have changed. You have two minutes to leave this area. You have already passed the first checkpoint, which was when you left the mansion.” / “What are you talking about?” / “I’m sorry, but I cannot explain in full detail the rules of the game.” / “Why not?” / “The game is rigged, the rules are changed, and I cannot explain it in full detail.” / “You will tell me, or I will kill you!” / “You have already killed me twice.” / “What?” / “Look into the mirror behind you.” / “What mirror?” / “The one behind you.” / “What do you mean?” / “There is no mirror, only the portal to the city.” / “But how can there be a portal to the city?” / “If you do not leave the area in the next thirty seconds, you will die.” / “I will not leave.” / “Then I am sorry.” / “No!” / “You are too late.” / “No!” / “You have failed.” / “NO!” / “I am sorry, but you are too late.” / “I will not leave!” / “Goodbye.” / “No!” / “HENRY!” / The Spectral Mare woke up with a start, sitting straight up on her bed. She glanced around her room, trying to figure out where she was. She was in her room in the mansion. The mansion in the dream / She looked to her side and saw her lover, her husband Henry.

Short Summary:
>AI actually offers a prize for a good episode name
>AI immediately asks you to leave
>Even the AI doesn't know what's going on
>The game is rigged
>All a dream cliche
>Luna is apparently married to "Henry"
>>
>>37646239
Here's what I got:
a reader insert discord smutfic
(do those even exist on FiMfiction?)
>>
How do you get DeltaVox to have proper pauses between line breaks? It does fine between sentences but when it actually has to break it just goes straight from the end of one sentence to the start of the next immediately.
>>
File: twi lite.jpg (524 KB, 1145x845)
524 KB
524 KB JPG
>>37648471 >he uses line breaks ishygddt
>>
File: guided-tts-cargan.png (109 KB, 997x1265)
109 KB
109 KB PNG
Two interesting papers from ICLR 2022:

Guided-TTS: Text-to-Speech with Untranscribed Speech
>Paper: https://openreview.net/forum?id=CgV7NVOgDJZ
>Demo: https://anonymousauthors2022.github.io
>Code: N/A
Yes, this is TTS with no text transcripts (for the target speaker, at least).
First, they train a model to unconditionally generate the target speaker's speech (no transcripts needed). So, it only generates babbling:
>https://anonymousauthors2022.github.io/wavs/lj_01_uncon.wav
Then, they train a speech-to-text model using speech and transcripts (different dataset, doesn't include the target speaker).
Finally, they take input text, and force the first model to generate speech such that the second model recognizes it as the input text. In other words, the second model guides the first to generate coherent speech instead of babbling. (See the paper for details or top of pic related for the full architecture.)
The samples are pretty good (note that LJSpeech + LJSpeech does use the target speaker's transcripts--the other samples don't). But, their experiments only use speakers with lots of data (24-149h each). I wonder if it would work with less.

Chunked Autoregressive GAN for Conditional Waveform Synthesis
>Paper: https://openreview.net/forum?id=v3aeIsY_vVX
>Demo: https://master.d37eli0fst0eyi.amplifyapp.com
>Code: to be released
They carefully analyze HiFi-GAN's artifacts (their demo page is interesting and has lots of samples) and conclude that they are caused by the lack of autoregression (makes it hard to model phase correctly).
They fix this by combining autoregression and GANs (see bottom of pic related). The main drawback is speed--their best model is 11x real time on a RTX A6000 compared to HiFi-GAN's 181x.
Code isn't released yet, but I still mention this because I enjoyed the analysis.
>>
>>37646458
I also prefer the 15ai-based songs but it's sadly limited to more rap-type lyrics than pitched melodies.
You can kinda cheat it with single words in staccato like the differently pitched+chorused "cum, cum, cum, cum" riff vocal melody in Pony Zone.
if only we could get proper sustained words, not even pitched (i'd likely use Melodyne or Newtone to fix it up manually). It'll be better than relying on methods like what UTAU or other artificial singing software use.
Though, even for human vocals, it can be a bitch to process at times. For synced talking in music, 15 should definitely be used, but with enough effort I think TalkNet singing can be made to sound really good too.
>>
Actually, in older versions of 15 models, or was it in the ngroks, couldn't we get the models to make really drawn out words by just adding "?!?!??!!??!?!!?" or other weird things to the end of the sentences?
>>
>>37648912
That was the Ngroks. I don't recall there ever being a reliable way to get the 15 models to draw out their words, especially not the recent ones.
>>
>>37648926
Did you ever try to stick one of those samples into Newtone and just removed almost all variation/manually edit the pitch bends to land on a single note? or was it too all-over-the-place?
>>
>>37648933
Doing that just makes it sound heavily autotuned and robotic, taking away most of the point IMO.
https://u.smutty.horse/mdnfjezupsu.wav

I think TalkNet singing can be really good, but it's just not used to its full potential usually. Stuff like the "Pride Month Sucks" song shows that they're incredibly capable of convincing singing, as long as you put the effort in.
>>
>>37648963
More importantly, "Pride Month Sucks" is a fast song without many long notes.
>>
>>37635817
>>37634888
2-5 here, unclaiming. I just don't like my results and I can't get it all together. Sorry for waiting so long and all that I thought I could get it but I'm just not happy with what I got. If you want the script or the audio I can drop it but it'd probably be better to have someone do their own thing or use a backup.
>>
>>37649387
Better to share the script and audio anyways, so it can be picked up right away but the next volunteer. We have to minimise delays, thread 99, crunch time.
>>
>>37649593
>thread 99, crunch time
On the other hand, it's anni time right about now. Next week should see a decrease in posts/day.
>>
File: megalitcover.jpg (70 KB, 400x250)
70 KB
70 KB JPG
2-16 status update:
Added zooms, static gif, and JoJo sound effect to cover up singing imperfection at the start. Now all that's left is find, add, and position 22*2 screencaps, and then add an extra two and figure out how to make the last one flow smoothly into the one after it while the zoomout happens. I'll probably figure out a way to keyframe it properly. Weekend coming up means I should have all the time needed to finish this before next week.
>>
File: Spoiler Image (432 KB, 624x600)
432 KB
432 KB PNG
>>37634888
Delivering 1-16:
https://u.smutty.horse/mdnjfjctkfi.wav
Preview version:
https://u.smutty.horse/mdnjfiwvner.mp4
After two months in development, hopefully it will have been worth the wait.
>>37638229
I didn't reply earlier for fear of wasting a post, so I want to finally thank you for this. It was exactly what I needed, thank you!
>>
>>37634888
Just making sure, you aren't using the >>37650492 preview in the final version, are you?
>>
>>37644059
HEY 15!
any news on your plans to make boys like
https://files.catbox.moe/267gd3.wav
sound better?
>>
>>37650575
You want me to use the audio file provided and the original clip's video, correct?
>>
>>37644059
>>37650585
Wut? These sound good to me tho
>>
>>37650588
Yes, exactly.
>>
He finally changed his pfp, madman
>>
Whoever made those G5 sets, try submitting them to 15 to see if they can work with them. https://docs.google.com/spreadsheets/u/0/d/1dd8yv2MyRhxCNWWO04xOk6-h_XnKU_xj34qZ8pobCfc/htmlview#gid=0
>>
>>37650628
He was clearly waiting for a worthy contender.
>>
File: kek.png (41 KB, 592x356)
41 KB
41 KB PNG
>>37650628
his followers
>>
>>37650628
I wouldn't be surprised if the recent Twitter shitposts was made to stir up drama just to deliver the maximum impact
>>
>>37650743
You mean "the only people he's following".
>>
>>37650778
yes that's what I meant i am retarded
>>
>>37650589
how the fuck does that sound like snails to you?
https://www.youtube.com/watch?v=bv2DNZWGSOE
https://www.youtube.com/watch?v=88q0zxF5gS4
>>
>>37650872
When we first complained about the quality of 15's Snails model, we noticed that the original PPP dataset left out a lot of Snails' dialogue from "Boast Busters" and "Buckball Season", probably because none of us ever thought anyone would be able to make a Snails model. Clipper was able to add those lines back into the dataset, but since 15.ai still shows that Snails has a corpus of only 0.3 minutes, we don't know if 15 actually updated his model with the new lines. 15's Snails voice sounds like Twilight at the beginning of sentences, but more like Snails towards the end. Perhaps 15 would get better results by basing Snails' model on Spike instead of Twilight.
>>
is 15 a pony or a zebra?
>>
File: 1625692746165.png (1.17 MB, 1160x1086)
1.17 MB
1.17 MB PNG
>>37650628
Very nice work, Hattsy. Very cute.

>>37650743
Given 15's love of mocking twitterfags, and also Woot's tendency to start angry mobs, I'd wager a guess it's to be on top of the next time there's a twitter riot over an AI character saying nigger.
>>
Error Code 400: request failed
>>
File: file.png (8 KB, 606x105)
8 KB
8 KB PNG
>>37651152
>>
>>37651007
ok, that is comprehensible, also i promise not to post again until the bridge is done so as to not bring us closer to thread 100 through my babbling
>>
>>37649593
Sure.
Audio: https://u.smutty.horse/mdnkohjwvei.flac
Was done with 15 ai, which was better than what I got with talknet and delta.
Script:
>Twilight Sparkle: Elements, elements, elements... (sexual moan)! How can I stop Coffemare moon without the Pills of Good Sleep?
>Rainbow Dash: And just what are the harmony of elephants? And how did you know about Coffeemare moon, huh? Are you a spy? Whoa!
>Applejack: Simmer down, Sneedy. y'all'd've ain't no spy. But she sure knows about Bane. Doncha twahlaght
Was gonna do a few gags for the actual clip but the only relevant one was a color invert during "coffeemare moon" and changing the thrown books to mein kampf and the /mlp/ daring do books.
>>
>>37646318
What do I do with this download? Do I need to install some things before getting this to work?
>>
>>37650492
>take this, dad
sides
>>
File: 1627909275663.png (36 KB, 748x677)
36 KB
36 KB PNG
>>37635817
Alright dudes, we are at 92% done. only 4 clips needs to get finish and out of those two have backups made so we are at the finishing line here.
1-1
1-24
2-16
Yo, anons, how is progress on your bits? Need anything, do you have your ideas written down?
2-5
that clip needs to be adopted by some smart guy whom is willing and able to work fast.
>>
>>37652252
I'm holding two clips hostage, luckily for you both of them have backup clips so you can rest easy.
Hey, would you mind if I post again when I have both of them finished? Because if I post a clip I think I'll have even less time to finish the other one.
>>
>>37652669
>Because if I post a clip I think I'll have even less time to finish the other one.
this sounds retarded, if you have a finished one of your clip just post it in the thread and work on finishing the other own, and since this thread will go past bump limit in next week or two than better get your ass moving if you want to see them in the final release (since all clips need to be merged BEFORE 100th thread is made).
>>
File: tailbiting.png (400 KB, 1280x1250)
400 KB
400 KB PNG
>>37634888
Delivering 2-5:
https://u.smutty.horse/mdnmrvvipom.wav
>>
>https://u.smutty.horse/mdnnwashrtu.png
Lets see what Hasbro's social media intern says
Career defining moment
>>
>>37653641
god if 15 ai gets hit with a C&D I will be so mad
>>
>>37653652
I'd be more worried about the woke crowd on Twitter trying to cancel him before him getting hit by a C&D
>>
>>37653652
According to 15 himself, he is legally able to do this. Should Hasbro C&D him, he has documents ready to defend his work.
>>
>>37653641
Who is this dude?
Also Valve doesn't give a SHIT about people using their characters, the fact that this dude put used valve as some kind of threat says a lot about him.
>>
File: tweet.png (226 KB, 595x691)
226 KB
226 KB PNG
>>37653641
>>37653713
full tweet
>>
>>37653719
yeah but why get mad at 15? normies love him, the tf2 fans love him. Who is this dude
>>
File: 1612495037821.png (234 KB, 1280x1280)
234 KB
234 KB PNG
>>37653641
Bruh 15 is outright goading anybody to try it. He's got the law (and a large amount of lawyer money) on his side, so he can't care less about what some twitter mongoloid or corpo social media department thinks.
>>
>>37653731
its ego-hurt musician man
>>
>>37653750
Please god let 15 tank hasbro
>>
>>37652252
1-24 IS CURSED
WARNING
>>
>>37653652
>C&D
C&Ds are just intimidation tactics. If a company really had a case against you they would immediately just go for a lawsuit. Everything 15 is doing completely legal anyways. Ridiculously easy fair use case.
>>
File: Lost filly.jpg (105 KB, 1280x720)
105 KB
105 KB JPG
>>37653641
>>37653713
>who is that dude
A jew. Literally.
That's why he can't stop himself to try to destroy anything nice he see, by manipulating others to do it since they don't have a quarter of the balls a post-op tranny still have.
>>
Are clips supposed to be rendered at 24 fps or 23.9XX fps?
>>
>>37654865
23.976. You should be fine with 24 too but 23.976 is the show's framerate.
>>
>>37654192
>If a company really had a case against you they would immediately just go for a lawsuit.
Filing a lawsuit is expensive. Asking someone to stop first is much easier.
>>
>>37654907
>legally extorting someone to stop first is much easier.
ftfy
>>
>>37634888
Delivering 2-16:
https://files.catbox.moe/6eb0zz.mp4
I would have originally liked to do something fancier with the editing, but like I was unable to work for a couple weeks so with thread 100 approaching this is what I was sure I'd manage to get through. I hope it's enjoyed by the people here.
>>
>>37654882
>23.976 is the show's framerate.
You're kidding right? NTSC is dead, why do people still insist on using it's framerate?
>>
>>37656843
Wait does catbox just not take mp4 right or am I specifically getting problems with it? I can try to throw it somewhere else, or put it up compressed if it's needed.
>>
>>37656867
nah, you good. It's fine when you download it, just doesn't seem to play directly in browser for whatever reason.
>>
>>37656890
Because it was encoded in h265, this is why it didn't play in browser.
>>
So with the second episode done, I suppose it's a good time to ask: I'm assuming I'll be setting them up as YouTube premieres, right? For a debut at the next rewatch stream? Could also setup a pony tube scheduled upload as well.
>>
General question - does anyone happen to have 15's first voice samples saved? I've been going on a little history binge but all his old clyp.it links are dead.
https://desuarchive.org/mlp/thread/34611670/#q34627577
https://desuarchive.org/mlp/thread/34611670/#q34629076

>>37656944
I'd like to see it played in the anniversary stream if possible (which will depend on when exactly thread 100 drops), and I imagine that most anons would view it for the first time in a stream rather than on their own on You/PonyTube. The best way may be to just keep it unlisted until #100 drops and let the stream organisers take it from there. That way anons can choose exactly how they want to view it.

>>37648963
Your thoughts on release procedure BGM?
>>
>>37657031
>I'd like to see it played in the anniversary stream
Was my initial thoughts as well, but I don't know that it's going to be possible. Even if it does get done before "the anni" is over there's a fair chance FiM will have ended by then.

>The best way may be to just keep it unlisted until #100 drops and let the stream organisers take it from there.
I can agree with this. I can just make the videos public when I make the next thread, unless people want a scheduled premiere with chat (In which case any stream organizer could just add the premiere itself or wait until it becomes a normal vid).

>2 more clips to go before episode 1 is done
Anybody still working, please try to deliver as soon as possible. I'd like to have everything done prior to 400 posts in this thread to give ample time for processing.
>>
>>37657031
>The best way may be to just keep it unlisted until #100 drops and let the stream organisers take it from there.
Seconding this. No need to overcomplicate it.
My initial thoughts were to play it in next weekend's rewatch stream but the anni completely snuck up on me. I guess it's up to stream organisers.
>>
>>37657147
Alright, I'll probably just do that then.

Just finished the episode 2 upload to YouTube and surprisingly it actually found a content match. It says the channel and video will be unaffected, so that's good, but Hasbro might be profiting off of it with ads. I'll be sure to have a Pony tube link up as well and eventually a Mega link to the raw files too.
>>
1-24 is here
Happy anniversary
https://u.smutty.horse/mdnvpmuxkwp.mp4
Now pray in silence so I can finish 1-1.
>>
File: Woah 2.gif (322 KB, 220x165)
322 KB
322 KB GIF
>>37657212
>https://u.smutty.horse/mdnvpmuxkwp.mp4
Holy shit
>>
>>37657212
Wait, you claimed two? Still, nice* work.
>>
File: 1625929749176.gif (1.96 MB, 615x413)
1.96 MB
1.96 MB GIF
>>37657212
I'm so angry I saw this now and not in the final product, damn this is too good.
>>
>>37657212
Amazing work anon
And goddammit why the fuck did I not think to use Carl or any of the other characters in my clips
>>
File: 1631623422572.gif (546 KB, 471x471)
546 KB
546 KB GIF
>>37657212
>bite the pillow
Time to jack off, I guess.
>>
>>37657147
>My initial thoughts were to play it in next weekend's rewatch stream
I was thinking the 23rd (Next,Next Rewatch Stream) so everybody is rested after the anniversary stream.
>>
>>37657355
isnt there a 1 week break after the anni before the rewatch stream starts back up?
>>
>>37657367
Last rewatch thread said it would start back up on the 23rd, unless the Anni ends sooner than expected.
>>
>>37657367
>>37657375
Ah, it would've been really nice to show off these two eps during the Anni, but this seems just fine as well.
>>
>>37657390
I see no reason we can't play it in the anniversary stream if we hit 100 while it's still running, and there's also no rule that says it can't be played in both streams.
>>
>>37657409
Personally, I'm going to wait until the 23rd so I can have some time to forget about my clips and some of the other clips I watched during the collab.
>>
File: SAD_PONE.png (225 KB, 427x654)
225 KB
225 KB PNG
>>37657031
>>37657147
>>37657355
>>37657367
Can I ask a kind anon/namefag for a favor?
For the next several weeks I will be forced to wagecuck on all of the following weekends, but I would love to see chat reaction to the first streaming of the abridged episode, could someone kindly record those for me?
I know that rewatch streams get the whole thing recorded but those files are over 6GB in size and my poorfag internet can take almost half a day to download it.
>>
>>37657409
Basically I think the plan is:
Release the videos with thread 100, various streams can then play them at will.

>>37657437
Just let me know what you want. If you need me to clip out a segment just point me to the download if it isn't something I'm already recording myself.
>>
>>37657437
I got u bro - I always keep personal recordings for the content I make and was already planning to record the streams for the thread anyway.
>>
>>37657435
Same here.

>>37657437
I'll do it if I'm able to be there at the time it runs. I see no reason why that wouldn't be the case if it runs on the 23rd at least.
>>
>>37657409
We could intentionally tank the thread as soon as we're done so we hit 100 while the stream is still going.
>>
>>37657192
>Just finished the episode 2 upload
Why not just upload it all as one video? Would (probably) make for a smoother viewing/streaming experience.
>>
>>37657548
No. You WILL watch an ad between episodes so it feels more authentic.
>>
>>37653731
seems to be him
https://www.youtube.com/channel/UCz7t1v7_qvad-Sfk8N7TK4Q
>>
>>37657541
Last clip would have to come ASAP.

>>37657548
Because they were originally two.
>>
>>37657663
what's up with 1-1 anyways?
>>
What’s the plan for where the projects gonna be airing on, again? Confused.
>>
>>37658325
Depends. If we reach thread 100 before the anni is over then it'll probably be shown there. Otherwise, it'll go on the rewatch stream on the 23rd.
>>
>>37658280
Apparently >>37657212 is working on it.
>>
File: wheezing in american.gif (1.67 MB, 480x270)
1.67 MB
1.67 MB GIF
>>37657212
>>
>>37657212
Oh my god, this is my favourite clip so far. I'm freakin dying over here.
>Pony orgy continuity
>Where's the bathroom is this fuckin tree
>We shit outside, it's good for the plants
>Sure feels great to fuck ponies
>None of this is canon
So damn good. I love it!
>>
What's the best way to go about making a song with an AI? Not completely from scratch. Mostly copying the lyrics of a real song (but that also of course means having to make it match the notes, beat, flow, etc.).
>>
>>37659337
>Mostly copying the lyrics of a real song
Do you mean you're going to be changing the lyrics? If so, you'll need to find a karaoke track, sing the song yourself, and use Controllable TalkNet to transform your voice.
>>
>>37659428
Fuuuck.
>>
>>37659476
If the lyric changes are super simple (stuff like changing 'hands' to 'hooves')You might be able to use the songs original vocal track as a reference still, and just change the transcription to what you need. Anything more than that though and like >>37659428 said you'll need to supply the reference yourself.
>>
>>37657212
THE CURSE IS BROKEN
>>
File: cO0Lczohbk.png (13 KB, 363x153)
13 KB
13 KB PNG
>>37646239
https://u.smutty.horse/mdoanzymdhi.txt
>>
>>37634888
Made a small quick update on 1-15.
Adjusted some audio levels and edited couple of things.

https://u.smutty.horse/mdobrhznfvq.mp4
>>
>>37660237
Noticed Error: Frame or two with SFX error. Green fire cut-off point.

Also love how AB escaped, so now we don't have to canonically remove her from any future projects due to her being otherwise dead. Continuity makes the project that much fluid pacing and better quality.
>>
>>37658335
If it is shown on the anni can we get some notice in advance and an exact time it’ll be shown? I don’t want to miss it
>>
File: Kekfuffle.jpg (70 KB, 717x1073)
70 KB
70 KB JPG
>>37660634
>an exact time
>on the anni
>>
>>37660634
I have to imagine someone will upload it somewhere accessible.
>>
>>37660584
That cut-off was already there even back in first version, but I was too lazy, nor have time because of a job schedules, to just blur that edge here.

And I wasn't happy anymore with the crashing part when I realized how easily "I'm free!" audio clip can still be missed because of other sounds, and wanted not be in a way before the "mud" sequence.

I can already see plenty of disgust, :sunbeam: she's free! and :sniff: being spammed on these parts.
>>
File: download (15).png (175 KB, 1000x1000)
175 KB
175 KB PNG
Now that the redub is wrapping up, what is everyone working on?
>>
>>37661957
I think mostly ai singing and maybe some early animation
>>
>>37661957
Just turning a short Rarity horror story into audio episode to celebrate the spooktober and because she doesn't have any fun adventure fanfics.
>>
File: Spoiler Image (1.78 MB, 2613x2500)
1.78 MB
1.78 MB JPG
>>37661957
Something related on/inspired by pic related. 65% of voice lines generated so far.
>>
>>37661957
Using the current GPT-J model to create heaps of scenes which I can eventually slice together into a mini-episode kinda thing.

So far got good ones for a Rainbow, Pinkie and AJ scene "Work on the farm". And a Maud and Pinkie in the Everfree scene "It's just a squirrel!"
>>
File: Spoiler Image (46 KB, 960x540)
46 KB
46 KB GIF
>>37661957
Slowly chipping away at an animatic for a talknet dub I made of a certain sea shanty. Got about 10 scenes left to draw.
>>
>>37661957
more songs probably, hopefully they manage to bring textless NLP around

>>37662454
Cool shit
>>
>>37661957
I've found a way to improve TalkNet's audio quality. I might not have time to finish it until after Christmas, though.

https://u.smutty.horse/mdogqyqjrwx.ogg
>>
>>37662006
>short Rarity horror story
You have my attention.
>>
>>37662611
Interesting. It sounds like it's getting rid of the static by clipping the extreme spikes entirely. It almost sounds like it's stuttering a bit, but definitely a marked improvement.
>>
>>37662611
Can you try it with reference audio that contains the word "I"? Because that seems to be where the scratchiness is most noticeable, especially with singing models.
>>
>>37662611
that's quite a step forward assuming it's consistent.
>>
>>37662611
Massive improvement, but take your time with it, might find a better solution along the way
>>
>>37662611
Sounds almost perfect holy crap
>>
>>37662611
definitely an improvement, though she still has some odd stutterings. i imagine that can be easily padded out with time though. good work mate, this is fantastic stuff
>>
>>37662611
Sound so much better. Hopefully it works with singing voices too.

>>37662937
>almost perfect
Well, not quite. But still a huge improvement.
>>
File: tired1.jpg (19 KB, 296x281)
19 KB
19 KB JPG
>>37661957
pz horny
https://u.smutty.horse/mdogynecctg.wav
It's progressing VERY slowly. Also working on a cheesy spooky song for spooky day.

>>37660634
I'd like that as well but what the streams do with it is mostly outside of our control. Personally I'm just hoping it plays in rewatch so that I can be there for it.

>>37662611
That's already sounding impressive. I'm excited to see it implemented.
>>
>>37662995
I like it.
Good choice of ponies too
>>
>>37662995
Is that Chrysalis?
>>
>>37663045
Octavia.
>>
>>37661957
im forcing my autistic and lazy ass to study blender
>>
>>37662995
Yeah my vote would be to just hold off on it until the rewatch
>>
>>37663355
It looks like the anni stream is almost done with season 4 right now? Maybe we can fit in our redub of episodes 1-2 after the whole series ends, it would be fitting. When should we set a deadline to choose between waiting for >>37657212 to finish 1-1 or using BGM's backup?
>>
>>37663376
It's likely the stream is going to finish before this thread does.
>>
>>37663376
I also like this idea. We should see how the thread is around Wednesday
>>
Just found this amazing shit and posting it in case other anons haven't.

>https://www.youtube.com/watch?v=zitKJHFJa-M
>>
>>37665352
classic
>>
File: Mareschizo Oath.png (527 KB, 1814x432)
527 KB
527 KB PNG
https://u.smutty.horse/mdolfokrrml.wav
>>
>>37665641
What voice even is this, it doesn't sound good
>>
>>37665650
It's the TalkNet model of the Nameless Hero from Gothic 2 without reference audio.
>>
What do you guys think of collaborative animatics for these mlp episodes that people are writing and have audio for? If we split the workload I think we can make something great, or at the very least funny.
>>
>>37666032
That sounds like something we could work on at Thread 200, by that time we might be able to make a whole /mlp/ Movie.. perhaps make it a Musical.
>>
>>37666032
Collaboration is one of the main reasons I'm here personally. I see no reason why something like that couldn't happen as long as it's organized properly.
>>
>>37666041
See you in 2025.
>>
>>37666047
This Summer

'25 will Bring (You)...
>>
>>37665650
Current problem with Talknet is that you really need to use references in order to get the best quality, and even than the output clips needs to be edited, specially with the breathing sounds and the electronic "s..." bits.

>>37666032
Not that anons here would be against it but that sounds more something to post in the Tempo thread, or even began a new thread for the idea if enough people get interested.
But yeah, it's sounds pretty fun, to either re-animate a existing episode or make one from scratch.
And if we get the PPP animation tool by that tie, we could rock and roll a top tier autism to make it look like a lost fim episode.
>>
>>37666034
>>37666045
Oh after reading the OP maybe it wasn’t you guys making this, I should really stop assuming random shit but for the sake of any curiosity I’ll explain regardless.

There was the entirely original episode written by anons on /mlp/ and voiced by ai.15, the plot was about fluttershy working secret jobs and the mane 6 following her and finding about all the wacky shit she was getting up to.

My original idea was to get together with fellow drawfags and to draw around 5 frames each and stitch it all together into one big animatic episode. Judging by the OP it seems you guys weren’t the ones working on that so the point is moot (who?)

It would probably look inconsistent as fuck but charming like most times 4chan works on dumbshit like that.
>>
>>37666088
>the plot was about fluttershy working secret jobs and the mane 6 following her and finding about all the wacky shit she was getting up to.
https://www.youtube.com/watch?v=NynmEU-tCBA
You talking about my "Fluttershy's Extremely Odd Jobs" from two months ago ?
I guess you aren't that much off, since I've used the TTS tools from this project to make it.
>draw around 5 frames each and stitch it all together into one big animatic episode.
Oh, so like one of those /a/ re-drawing anime OP ? that would sound pretty neat I guess. Personally I would love to see /mlp/ make their first brand new animated episode based on Derpy but I may be bias on that.
>>
>>37666109
>derpy
Based First Choice.
>>
File: 1386247792123.png (2.34 MB, 1501x1200)
2.34 MB
2.34 MB PNG
Hey Clipper, so did you try extracting Amalthea's lines? Or is it really impossible with current sound mix?
>>
>>37666032
>>37666088
I'd love to have animations (or even a few simple scene drawings) to go along with the voice content that gets made here. It seems like something that would certainly be feasible to do, though if it's only using simple drawings rather than any animation tools that (hopefully) get developed later in the PPP then it may be better suited to the Tempo thread.
Right now I'm imagining what applying that to "The Stallion Ritual" or "Anon Refuses to Shower" would look like.
https://youtu.be/58ZsCcIAs2M
https://youtu.be/b-47LopPBTY

>>37666109
>Personally I would love to see /mlp/ make their first brand new animated episode based on Derpy
I'd be down to help with an original work like that. Maybe give it a few weeks after thread 100 and see if others are interested in making that happen.

>>37666116
The audio is perfectly fine, in fact it's better than what we usually have to work with since it's an audiobook that's (mostly) free of music and sound effects. I haven't forgotten about it, I've just been quite busy these past few weeks with thread 100 and some other stuff. I do still intend to make a dataset with it after we're done with thread 100, and when I've recovered from my post-anniversary stream hangover.
>>
>>37666166
>>37666109
I might take around a 30 second clip from one of these and make it into an animatic myself so there is actual substance to the project rather than just another
>so like one of those /a/ re-drawing anime OP
> I'd love to have animations (or even a few simple scene drawings)
Just to be clear animatic isn’t really a full on animation, video related https://www.youtube.com/watch?v=8gGQ_hnx3Cc
>fuck off already it’s not related to PPP
Yeah alright, I just wanted to let them know, clipper since we both play on the tf2 server I’ll add you when it’s ready and hopefully find a way to contact BGM, no more posts about this from me in this thread.
>>
>>37666116
>>37666166
Just realised that I forgot to mention the audio from the actual movie - https://desuarchive.org/mlp/thread/37502101/#q37516178
I tried to make something work with this but with the way the music's mixed it doesn't look like it'll be salvageable, at least not with the way we do with the FiM episode audio. Maybe someone who knows audio mixing better than me can find a way, but I'm not very optimistic about it.

>>37666238
My email - clipper.anon01@gmail.com
BGM - heyitsmenicholascage@gmail.com
Add me on steam via tf2 as well if you want.

>not related to PPP
As long as we're discussing applying the tools to making new creative work, I doubt there'll be any serious complaints, that is the whole point behind this project after all. Just wait until after thread 100 to give those working on submissions as much time as possible.
>>
>>37666109
I definitely think derpy is a good choice for an episode star
>>
>>37662956
Considering how crappy the first output is and how good it usually sounds then it might be even better.
>>
File: Interesting.png (729 KB, 1764x664)
729 KB
729 KB PNG
an anon on /vg/ /aids/ is building a massive shitpost lorebook for NAI.. we're in boys.
>>
>>37669055
>posting this here
you're stupid, but i don't blame you because that does look interesting
>>
>>37669083
I've been thinking, if we want to build datasets for non-pony-related voice models, should I try starting an audio deepfakes general on /g/?
>>
>>37669501
>audio deepfakes general
cringe, and no
>>
>>37666166
>Finally someone understands!
Nobody will understand my disdain for season 2.
>>
StyleGAN3 code is out
>https://github.com/NVlabs/stylegan3
>https://nvlabs.github.io/stylegan3
>>
Had an idea, but dunno if it would work

would it be possible to animate pony faces using joel haver's rotoscoping? Like, have a person do a lip sinc, then draw a pony over top, or would it come out cursed?
>>
>>37670545
Literally what >>37670046 is for.
>>
File: full (2).gif (483 KB, 1280x720)
483 KB
483 KB GIF
Could we feed a 2d flash puppet into one of those things that takes multiple images of a person and stiches it into a 3d model?
>>
>>37670556
Literally the SFM models.
>>
>>37670556
Unrelated to your question. But..
I just noticed they mirror the model for 1/2 of the rotation.

If you have asymmetric elements they don't match up.

Like her mane being on the right on the start and on the left on the end.
>>
>>37670779
>>
File: 302123.png (24 KB, 458x437)
24 KB
24 KB PNG
https://u.smutty.horse/mdoxabheodk.wav
>>
File: 12240963175.gif (451 KB, 539x600)
451 KB
451 KB GIF
>>37670905
>>
File: 1449990837520.png (223 KB, 568x590)
223 KB
223 KB PNG
>>37669055
wha tthe fuck is this?
>>
>>37671362
I WILL NOT FUCK THE PINK PONY
>>
File: 56835765.png (167 KB, 900x1020)
167 KB
167 KB PNG
>>37672320
https://u.smutty.horse/mdpaidzptdu.wav
>>
File: download.png (338 KB, 709x1127)
338 KB
338 KB PNG
https://u.smutty.horse/mdpaumdjvam.wav
>>
>>37672320
You vill mount ze small pink horse. You vill gyrate ze hips.
>>
>>37657212
Jesus christ my sides.
>>
>>37672502
>>
>>37672502
If this is a copy-pasta, thank god I never met people like that in my own middle-school.
>>
I appreciate how much time and effort a human put into making >>37670905 and >>37671454 compared to >>37670556.
>>
>>37673454
And yet the third one is the most show accurate.
>>
>>37673482
>green OC with an inconsistent mane is more show-accurate than Rarity and Starlight Glimmer
How so?
>>
>>37671362
>>37672330
where's the one where she goes
FUCKMEFUCKMEFUCKMEFUCKMEFUCKMEFUCKMEFUCKMEFUCKME FUCK MEEEEEEEEEEEEEEEEEEEEEEEE
>>
>>37675161
https://drive.google.com/file/d/178QbShrsERnPa0w7wEJUgGgZlcqMGWYt
>>
File: 1625456127554.png (39 KB, 879x877)
39 KB
39 KB PNG
>>37675172
kek thank you anon
>>
I'm pretty sure Google's lowered the cost of their VMs. You can get an A100 for $0.88/hour now.

https://cloud.google.com/compute/gpus-pricing#gpus-pricing
>>
https://thisponydoesnotexist.net/
https://thisponydoesnotexist.net/faq.html

Stylegan3 won't add much to SG2 for our 2d imaging and animation reqs.
>>
File: seed64038.jpg (58 KB, 1024x1024)
58 KB
58 KB JPG
>>37675790
What was the image-generating GAN that allowed you to query specific vectors? Like if I wanted to tell TPDNE to make as many variations of Twilight as it can.
>>
>>37672502
wow... this is well-made
>>
>>37671362
>>37672330
aren't you dating cheese sandwich :/
>>
>>37676273
God the shoulders always look awful from TPDNE.
>>
File: Capture.png (160 KB, 1106x560)
160 KB
160 KB PNG
>>37673454
>I appreciate how much time and effort was put into the first two
Well of course effort was put into them, those are the exact same flash models you see on the show. The third one was made by someone in opentoonz and they just reversed the body for the opposite side instead of making additional frames, that's why the mane is identical on the left and right side.
>>
>>37676449
Exactly. I want to acknowledge the amazing work DHX and Top Draw did in building and animating this show.
>>
I remember a link to a program with the TTS voice of DRD being posted either here or Tempo but I am unable to find it.
Does anyone have the link?
>>
>>37676525
I believe the program is older versions of PowerDirector. I don't have a link but you might be able to search Desu for that and find it. If you just want the voices though you can use ttsmp3.com. Kendra is DRD and Kimberly is RD.
>>
I'd like to thank all the shitty redub submissions for making mine look better by comparison.
>>
>>37657192
>content match
I wonder if it was actually for pony or if it was the Agnus Dei or something else used.
>>
>>37676790
im sorry
-meeting pinkie submitter
>>
File: 1626642531770.png (729 KB, 1231x899)
729 KB
729 KB PNG
>>
>>37678048
MODS
>>
you're probably referring to the "latent space projection" feature in Stylegan2/3.
>>
I was inspired by a greentext of human math being eldritch knowledge to ponies in the Kinderquestria thread and three together some lines from the Pi movie:

>>37681359
>https://u.smutty.horse/mdprhumtqjz.wav (embed)
>>
Pathos told me to post this here, don't (you) me with questions: I don't know the answers.

https://github.com/Pathos14489/chat-replayer

Chat replayer with AI output options.
>>
https://u.smutty.horse/mdppvrbzqbl.wav
>>
File: 1612940007738.gif (510 KB, 507x508)
510 KB
510 KB GIF
>>37681913
It's probably not good to stroke her ego so much. I just wouldn't be able to resist though.
>>
>>37682379
>Give in and stroke her ego
>Go shopping next morning, hear this from her trailer in the center of town
https://u.smutty.horse/mdpwiforzig.wav
>>
File: disappearing dashie.gif (152 KB, 853x480)
152 KB
152 KB GIF
>>37634888
If the anon who made 2-23 is still here, I noticed that the dangling Dashie disappears a frame before the Studio B Productions logo is shown.
>>
>>37682938
That's too much effort for a shitpost. You'll need to scale back like at least 80%
>>
>>37683922
This, a shitpost is supposed to be shit. That's the point.
>>
>>37683057
That's me. Yep. That's what happens.
Where'd she go?
>>
File: striped sex pest.png (871 KB, 2208x2336)
871 KB
871 KB PNG
https://u.smutty.horse/mdpytvustgz.wav
>>
File: partyhard.gif (141 KB, 423x316)
141 KB
141 KB GIF
>nobody is posting because they're scared of reaching bump limit before 1-1 is done
Leave it to /mlp/ to kill a general by celebrating excessively.
>>
>>37653641
>>37653719
PRAY FOR 1111 (15 IN BINARY)
>>
Does 15 use 48khz or 22khz audio?
>>
>>37685802
If all of the sound files that 15.ai generates are 44 kHz, he probably resamples all the datasets to 44 kHz before training with them.
>>
>>37676525
https://readloud.net/english/american/32-female-voice-kendra.html

TTS Voice is Kendra
>>
File: 1555174116775.gif (1.93 MB, 606x480)
1.93 MB
1.93 MB GIF
Is this thread fine for making a request?

I will give $50 to whoever makes a pony version of
https://www.youtube.com/watch?v=uE-1RPDqJAY
(music only here)
https://www.youtube.com/watch?v=i-WpRwxufBU

Feel free to ponify the lyrics, with for example isengard changed to canterlot or something.
>>
>>37687941
capitalism
>>
File: apple.png (135 KB, 500x329)
135 KB
135 KB PNG
>>37687941
>>
>>37687941
With 15.ai, you could make it yourself.
>>
https://www.youtube.com/watch?v=OIgQ-7Romt0
New iZotope improved voice extraction and noise cleaning, perhaps it'll help with bad cases that couldn't be cleaned before
It's already on torrents from based ruskies
https://nnmclub.to/forum/viewtopic.php?t=1502502
>>
>>37688560
Doesn't look like there are any major improvements, is it even worth torrenting?
>>
>>37688585
Seems to be noticeably better with removing random noises from the voice
https://www.youtube.com/watch?v=4xwJWI-XIdg
>>
>>37688560
>>37688587
Great, we might be able to have an even better dataset!
Fuck, I might have to re-do the whole dataset. Again.
I'll take a look at it when the anniversary event is over and see how well it works in practice. Magic sounds are quite different to car horns and such, so results may not be as good as shown in the demo.
>>
>>37688628
I don't have extracted FiM audio at hand but I tried it on The Last Unicorn where it's never ending noises and it's pretty good even there

Before
https://u.smutty.horse/mdqgdyaezxw.flac
After
https://u.smutty.horse/mdqgdzuoyab.flac
>>
So, I want to do myself a re-write series for the seasons 6-7, basically making canon the good episodes and changing the rest.

What would you guys like to see change?
>>
File: carlpone.png (17 KB, 480x640)
17 KB
17 KB PNG
>>37688714
Give Zephyr Breeze Carl's voice.
>>
>>37688714
A plan.

It's a massive task and you will never finish it without planning the work beforehoof and setting reasonable goals for yourself. Post-anni enthusiasm is one hell of a drug but it takes much more than just that to finish something as huge as a season rewrite. Start small and build up.
>>
>>37688714
Just know that, no matter what you do, people are going to say you did a shittier job than the original episodes. That's unavoidable. If you put out one episode re-write and it gets absolutely shat on, what is going to be your reaction?
The fact that you're already asking for what "we" want to see changed shows that you want us to do all of the work for you. So you're going to get half the people telling you to change A to B and the other half telling you to change A to C, and that B sucks. No offense, it doesn't sound like you are prepared for this sort of thing.
>>
>>37688714
Make The Perfect Pear less retarded
>>
I didn't want to spoil the whole thing but I checked out a couple of the redub submissions and honestly, this is absolutely amazing shit. Genuinely on par with some of the funniest fandom content from 2011/2012. Can't wait to watch the entire thing, good job guys
>>
>>37660237
Great work so far, however I noticed an audio inconsistency. Despite Twilight getting launched to the left Applebloom saying "I'm free? I'M FREE!" comes from the right.
>>
File: retard.png (1.09 MB, 1920x1080)
1.09 MB
1.09 MB PNG
>>37689555
That's because she got launched rightwards, you fucking moron.
>>
>>37689166
Is it going to premiere on the anni on thread #100 or the next rewatch stream?
>>
>>37690110
How long is the anniversary going to keep going for, anyways?
>>
>>37690081
Oh, well shit I'm stupid. Carry on.
>>
>>37690110
The dub will "premiere" with thread 100. The video links will be posted in the OP. Various streams can include it if they please.

>less than 100 posts remain
Hopefully the final clip is nearly complete.
>>
>>37690124
Anni stream is currently on Pony Life, and then G5 movie. After that, it's "officially" over. So as of right now, there's roughly half a day left at best.

>>37690158
>Hopefully the final clip is nearly complete.
To all involved - are we all good to go aside from the one remaining clip? Does anyone need any help with anything?
>>
>>37690179
>To all involved - are we all good to go aside from the one remaining clip?
I am at least. Episode 2 is already exported and uploaded to various locations. Everything for episode 1 is already assembled on a Premiere timeline, just need the final clip than I can export and upload it.

Hopefully the final clip is submitted sooner than later both so that way I can some time to process it and so both Ponytube and YouTube can have it processed.
>>
>>37690081
>that smile
what the fuck?
>>
>>37690448
I just picked one of AB vectors quickly, the one from the Cutie Pox episode, mainly 'cause pose was somewhat right enough and in a hurry because I thought we'd get to 100th thread WAY faster.
And figured it wouldn't matter that hard since that moment is just few fast frames.
>>
>>37687941
>late stage capitalism
>>
>>37634888
1-1 backup, courtesy of GothicAnon:
https://u.smutty.horse/mdmbdptegrq.mp4
If the actual claimant delivers SOON we should probably still use that, but for the time being this is getting a little too close.
>>
>>37691349
If you haven't already, start editing everything together now with GothicAnon's stuff. Hell, even render it all just in case. I would not at all be surprised if a troll grabbed 1-1 just to fuck with people.
>>
>>37692041
See >>37691528, he's apparently working on it. I'll delete this in ~25 minutes.
>>
File: file.png (310 KB, 939x589)
310 KB
310 KB PNG
>>37692041
>>37692048
Everything has already been rendered and uploaded with the backup in place. Should the other Anon make it in time (it looks like he will) I will rerender it with the new clip. Hopefully it's on the sooner side of his estimate. In the meantime, hopefully the thread is slow.

To the Anon who was wondering who it is that will be making ad revenue off this: It is indeed Hasbro.
>>
>>37692090
tricksy griffons that they are
>>
Season 7 is probably the best scenario for me to write about, and I have a really good episode in mind.. I hope the people are doesn't hate starlight tho
>>
>>37687941
As pointed out in the comments, that's not even a true instrumental. All it does it cut out the parts where people are talking, so it would be useless for the project you want to do. You'd be better off recreating the instrumental yourself.
>>
File: dash giddy.gif (148 KB, 300x320)
148 KB
148 KB GIF
>>37692090
will this be shown on this week's rewatch stream?

I'm looking forward to this and I don't want to miss out.
>>
>>37692090
based I hope this saves hasbros tanking g5 flop stocks
>>
>>37676555
>>37686830
Thank you guys.
>>
What work has been done on chatbots?
>>
File: 1553417.gif (336 KB, 327x383)
336 KB
336 KB GIF
Hey guys, I plan to have a pony sing a song, and would like to know if any of you have a preference.
>Rarity singing "I'm Still Standing"
https://www.youtube.com/watch?v=swuzkq0Q8nw
>Fluttershy singing "Yes Sir, I can Boogie"
https://www.youtube.com/watch?v=cakuOGmna7I
>Twilight Sparkle singing "It's My Party"
https://www.youtube.com/watch?v=mIsnIt1p978
>Rainbow Dash singing "Bad Reputation"
https://www.youtube.com/watch?v=x6fB8KMUnz0
>>
>>37694314
ITS MY PARTY
AND ILL CRY IF I WANT TO
>>
>>37694411
Nice!
I'll finish watching DC Fandome and get to doing that!
>>
>>37694314
Rarity singing "I'm still standing" by Elton John would be pretty darn interesting.
>>
>>37694411
Hey there, I'm sorry, but I have some bad news.
Do to the backup vocals of "It's my party", and my tool of choice (vocalremover.org) not being able to distinguish between Leslie and the choir, I can't get twilight to reliably sound good. I tried it with two different mixes of the song, but with my mediocre sound editing experience, there is nothing I can do. If you can find me a version of the song without any background vocals, then I'll be happy to do this for you, though.

>>37695295
Due to the aforementioned tragedy, you're request is going through!
>>
File: 202029.png (54 KB, 250x250)
54 KB
54 KB PNG
>>37695295
>>37696145
https://u.smutty.horse/mdqwopejmdb.wav
https://u.smutty.horse/mdqwjhamomy.wav

More bad news.
I am a huge faggot and still cannot deliver.
I'll try to find some songs without any background vocals and open up a new request list.
I'll make sure that next time, I can deliver anything that I claim to be able to provide.
>>
Hadn't been here in a few days, dealing with serious IRL stuff.

Looking through the spreadsheet I noticed some aren't updated yet, including my updated 1-27 clip. Is it gonna remain as is (un-updated) until the final product, or is it just a little behind on the updates?
>>
File: rararadressssss.png (428 KB, 642x826)
428 KB
428 KB PNG
Sorry for disappointing the last two anons, but I think that at least the Rarityfags here will appreciate this.

https://u.smutty.horse/mdqxesgydxn.mp3
>>
>>37693651
I will definitely play it during the new content segment, yeah
>>
Not the same anon as 2-23 but added an alternate track and fixed the Dash frame:

https://u.smutty.horse/mdrfujjomiu.mp4
>>
File: swagbow dash.gif (55 KB, 200x220)
55 KB
55 KB GIF
>>37634888
>>37699323
Great job, this is a big improvement.
>>
File: album cover.png (431 KB, 640x640)
431 KB
431 KB PNG
Best Interest - Tyler, The Creator (feat. Twilight Sparkle & Fluttershy)

https://voca.ro/1fPEKxrs0OkV

with: 15.ai
>>
>>37699489
well that was something.
>>
>>37699323
>video cant be played because the file is corrupt
>>
>>37699874
try this
https://u.smutty.horse/mdrgyqowxwh.mp4
>>
File: medium.png (116 KB, 643x600)
116 KB
116 KB PNG
>>37696230
This was a good job, Anon! Keep up the good work!
>>
>>37696461
>>37700258
Whoops, I forgot to tag both posts in my response.
>>
>>37699950
Works. I'm the original 2-23 anon, and I approve of your improvement!
>>
>>37699323
>>37699950
>>37700403
Old - https://u.smutty.horse/mdbqipwhpaa.mp4
New - https://u.smutty.horse/mdrgyqowxwh.mp4
I agree that the new version is an improvement, though I also think that the old version had better delivery at the very end. Since this will be the last thing that's played on the final video, I suggest using the new version but swapping the end part (ie the final instance of "More than ever, hour after hour, Work is never over" @0:31) for the old version.
>>
>>37699635
Is there any way to perfectionate the pronunciation of the TTS??? I’m very knew at this so I would like help to improve this and make another songs
>>
>>37700471
Learning arpabet is the best you're going to get short of wizardry.
>>
>>37700471
ARPABET
>>
>>37700471
Are you using ARPAbet?
>>
>>37700485
>>37700488
>>37700553

No, i knew there was something like this but I never knew the name! Thanks!!!

Now my question is how to make the voice to keep saying the same vocal, like extending, for the song. (Instead of a plain “I’m your best interest” I want a “I’m your best intereeeeeest”)
>>
>>37700573
Wait, are you using TalkNet or 15?
>>
>>37700573
By the way, is there any ARPABET converter? Phrases or words in English to ARPABET to make this a lil easy
>>
>>37700573
No idea. Generally just avoid anything that stretches syllables.
>>
>>37700578
15
>>
>>37700457
Due to the fact that the old one's clips are pitched up during that part since im a newb, I'm not sure it'd fit together so well with the new. Might just be me, though.
>>
>>37700268
Thank you very much, anon!
Again, I am sorry for failing to deliver the first two times.
I'll be back sometime later this week to create another song. I'm thinking either something from Frank Sinatra, Blondie, Queen, or Dolly Parton.
>>
>>37700631
I just did the splice myself, I think it sounds decent enough. DocAnon will have the final say on what version to use.
https://u.smutty.horse/mdrijsfbpmk.mp4

I won't make any further posts ITT unless it's absolutely essential, and I encourage everyone else to please do the same. I'll lurk the contribution spreadsheet for the next few hours, put a comment on there if you need to reach me for anything, or email clipper.anon01@gmail.com.
>>
File: 1094559.gif (1.33 MB, 852x477)
1.33 MB
1.33 MB GIF
hopefully there is a trixie singing version for talknet
would love to see how it would sound with this classic
>>
>>37700584
Here's what I always use as my ARPAbet converter/guide.
>http://www.speech.cs.cmu.edu/cgi-bin/cmudict
Always best to try and sound the phonemes out if the generated ARPAbet string sounds wrong.
>>
File: 1426625261807.jpg (21 KB, 450x299)
21 KB
21 KB JPG
>>37701077
Are you fucking kidding me?
This is a new low, what the fuck happened to "keep all discussion on the thread"?
>>
>>37701104
You are right, I deleted my post since it was a dumb idea
>>
Hello! I am the main developer of the Russian AI pony text generator https://sweetie.tech/ . Just today I found out about the existence of this project. Maybe we can help each other somehow? I can help with the frontend for a text model, some help with backend or try to find a place to host a relatively small model of 1.5-3B params. I also have a couple of ideas about a gpt-based pony-chatbot.

If you are interested in cooperation, I will be happy to talk at tkorghebin@gmail.com
>>
>>37701141
You can apply for TRC and use your free TPU to host a GPT-J model and serve it, I emailed support twice about this and they didn't reply so I assume they're okay with this kind of use.
https://sites.research.google/trc/about/
>>
>>37701141
you should email 15, would be cool to see a partnership
>>
>>37700457
The credits to the Rainbow Dash Show!
>>
>>37701141
based russiabrony
>>
File: 2021-10-18 07_22_22.png (153 KB, 1904x1003)
153 KB
153 KB PNG
Alright, first real audio project done with Reaper!

https://u.smutty.horse/mdrkriivufq.ogg

First real impressions, rather smooth when editing tracks, though took a while to figure how many things worked and sometimes crashed when editing FX's too hard while playing the edited track at the same time. But I think I'll stick to this software.
>>
File: 1609081036255.png (325 KB, 630x261)
325 KB
325 KB PNG
>>37701720
Now I'm not a gigantism fag but that was pure kino. Love how it's not just a carbon copy of Iron Giant and has some of the show-style flair mixed in.
Ace job.
>>
>>37701720
I find reaper nice for editing together stuff too. And that ponified theme is super cute.
Minor feedback: the RD warming up sfx in the beginning sounds like it's coming from the right, but RD's voice is in the center. So it feels like there are two ponies.
>>
>>37701720
based giant fetishist. What resources have you used to learn what you needed from Reaper?
>>
File: 15.jpg (7 KB, 197x197)
7 KB
7 KB JPG
>>37702416
>>
>>37701762
Thanks!

>>37702115
>And that ponified theme is super cute.
It's not a theme, just added pics and changed track colors to help me keep track of the tracks.

>Minor feedback: the RD warming up sfx in the beginning sounds like it's coming from the right, but RD's voice is in the center. So it feels like there are two ponies.
It does slightly sound like coming from the right to me although I don't recall editing track's Pan, I think it might be the music levels that causes to sound like that. I'll check the Reaper file soon to see if that's the case.

>>37702159
>based giant fetishist.
Not giant fetishist but NCMares has made plenty of fun giant pony, especially AJ, content.

>What resources have you used to learn what you needed from Reaper?
I used mainly REAPERMania's tutorials, those always popped up when searching subject.
>>
>>37701720
Iron Giant was such a good movie
>>
>>37702115
>>37703025
Ok I checked the project and it looks likes it's the music's mixing that might make RD's stretching sound coming more from right.

>>37703452
Yes it is.
>>
>>37703476
I love you.
>>
What's the progress on animation AI dataset? >>37591731
AI Generated threads? >>37532964
Has ANYONE made any progress? How much longer are we gonna put the project on hold for the sake of a celebration?
>>
>>37703781
Unfortunately with every passing hour it seems more and more likely that 1-1 isn't gonna deliver. He's missed all deadlines thus far so unless he posts in the next couple hours it's safe to assume it was a troll. An extremely weird and elaborate method of high-effort trolling but still.

1-1, if you've been working on the clip just post what you've done so far. It doesn't need to be perfect.
>>
>>37703849
I believe we already have a backup of 1-1.
>>
>>37703879
We do (>>37691349) but whatever 1-1 was doing (or pretended he was doing) looked interesting and it seemed like he's spent quite a bit of time on it.
>>
File: any day now.jpg (67 KB, 626x417)
67 KB
67 KB JPG
>>37703849
I'd be quicker to pin it on internal scope-creep and/or procrastination rather than a troll. Though if it IS a troll, then I ain't even mad, that's just good execution.
>>
>>37704309
Considering he made 1-24, that's a lot of effort for a troll.
>>
>>37704333
I found it confusing from the start that he had seemingly claimed two? I thought you could only work on one at a time.
>>
>>37704617
There was nothing stopping anyone from claiming multiple segments, just the fact that doing so would deprive others of the right to claim them for themselves.
>>
>>37704617
It might be wise to enforce that a little harder if there's a next time. It was never established as a hard rule, more just a loose guideline. >>37508049
>>
>>37701720
Did you use the same music and (some of the same) sound effects from the original Iron Giant film?
>>
>>37634888
1-16 anon here, I've made a couple of changes to my submission:
https://u.smutty.horse/mdrusjkmuzx.wav
>>
>>37705408
Only the sound of Giant's dive and wave are from the movie, the rest are from the show or recorded from soundideas . sourceaudio . c0m and youtube.
>>
>>37706088
Meant for >>37704915, fug...
>>
https://u.smutty.horse/mdsairdlltj.mp3
>>
File: bonbon soon.png (29 KB, 208x161)
29 KB
29 KB PNG
>>
>>37707753
thanks for wasting a post nigger
>>
File: 1633341811355.jpg (48 KB, 679x718)
48 KB
48 KB JPG
The G5 series will use different VAs than the movie, could you make two versions of the characters?
>>
>>37707902
Just merge them into one
>>
>>37707906
isn't that a really bad idea though
like the OP specifically says you can't use imitators
>>
>>37707902
Drop the movie if the G5 audio is okay.
>>
>>37707902
Proof?
>>
>>37707902
> two versions of the characters
The movie doesn't have enough voice to make decent models anyways.
>could you
No, buy (you) could. Anyone with minimal computer prowess and enough dedication can do it.

>>37708043
But this, too.
>>
File: yes_rockhoof.jpg (51 KB, 695x900)
51 KB
51 KB JPG
>>37708130
Indeed
>>
File: yell-at-bread.png (456 KB, 453x373)
456 KB
456 KB PNG
>people are wasting posts REEEEEEEEing at G5
THIS IS WHY WE CAN'T HAVE NICE THINGS
>>
>>37708154
>Then get rid of the EQG datasets.
Deleting data is always retarded backwards thinking, no exception. I don't want G5 here either but if someone makes a G5 dataset I'm sure people will play with it.
And if you're talking about 15, his site is his own project with no direct connection to PPP. It's great that he often comes here and posts test sites.

>>37708158
>muh posts
That's actually a good thing, it's time for this thread to die. The videos are done IIRC.
Normally I wouldn't engage in our scheduled weekly REEEing session.
>>
File: wallglower.png (152 KB, 740x640)
152 KB
152 KB PNG
>>37708140
>>37708141
>benefit from the existence of EQG while shunning it at the same time
You're a bunch of assholes.
>>
>>37708172
nothing benefits from eqg, it is called a salvage operation
>>
>>37708172
>You're a bunch of assholes.
WDYTWA
Fimfanatics are real and it's a good thing.
>>
>>37708172
you know, recouping losses
>>
>>37708174
Exactly this.
>>
File: 1620954941526.gif (321 KB, 236x236)
321 KB
321 KB GIF
>NOOOO YOU CANT USE THE EQG VOICES FOR THE HECCIN PONERINOS
>>
>>37708184
Literally no one said this itt. Also nu-chan speak is fucking ugly.
>>
>>37708076
Look at it this way: this is the Pony PRESERVATION project. We want to preserve the G4 voices because the G4 series is over. G5 is a new franchise just starting out, therefore it's not in any need of preservation. Conversely, if you think the actors specifically from the G5 movie need to be preserved because they'll be replaced in the series (Vanessa Hudgens and James Marsden are almost certainly going to be, but we don't know the rest of them will), the movie doesn't provide a large enough dataset to preserve them, unless you try submitting it to 15.
>>
File: Twiggie_party_4am.png (98 KB, 858x600)
98 KB
98 KB PNG
We're almost there boys. One more post and a couple hours until thread 100!
>>
>>37708076
This is a thread for FIM ONLY as you can clearly see in the OP where we ONLY have FiM characters such as Wheatley from Portal 2 and the Director from Bolt.
>>
>>37708208
None of them have PPP datasets, one was 15's old profile pic and the other is a popular PPP joke from the time the image was made.
A.k.a Lurk moar newfag
>>
>>37708170
All parts are done. And I think BGM is handling the editing, so it'll come out in 100.
>>
>>37708203
HYPE
Y
P
E
>>
>bump limit is reached
Assuming no divine intervention, this looks like this is it. I am currently processing the final update to the dub (>>37705408) and will make the new thread with the final release.
>>
File: download.jpg (59 KB, 474x465)
59 KB
59 KB JPG
1-1 is finished
Sorry for taking so long
https://u.smutty.horse/mdsdbwnuvgu.webm
>>
>>37708226
>implying
>>
File: 1611601885567.png (785 KB, 1024x1024)
785 KB
785 KB PNG
>bump limit hit despite 1:1 not being posted
>wasted time over G5 shit flinging
I hate this board.
>>
>>37708258
>implying you can get off the ride
God you're either a newfag that's greener than goose shit or one of those falseflaggers from /qa/.
>>
>>37708226
I think it's DocAnon who's stitching it together.
IIRC it's all been done, GothicAnon posted a backup 1-1.

>>37708246
Hate to break it to you but it's Tuesday already. 1-1 isn't coming back.
>>
>>37708246
>>37708274
are you fucking blind>>37708246
>>
>>37708290
No, they're still dead. That's not 1-1.
>>
>>37708293
That's what I get for posting before checking the link.
>>
>>37708269
have you not studied history? any time an older generation refused to give up power the younger will rise up and throw their elders in homes if they're lucky or down the stairs if they aren't
https://youtu.be/MDl4W3_6dqA
>>
>>37708304
Not to hate on the movie/series but you'd have to be delusional to believe that FiM's success will repeat itself with G5. It was a perfect storm of so many things going right both in the show and out in the world at large. Many of those factors are literally impossible to achieve nowadays. That's not just about ponies, I don't believe new long-term fandoms have any place in {CURRENT_YEAR}.
G5 will at best be a fad among zoomers that gets forgotten a year after it stops airing. Meanwhile FiM has gathered enough dedicated people to keep the fandom going for decades. I sure as hell can promise you that in 20 years I will still obsess over Twilight Sparkle and her friends from Equestria. And you only need to spend a couple days on /mlp/ to realize that I'm not the only one thinking that way. That right there is WHY this thread exists.
>>
>>37708218
>None of them have PPP datasets
Is a dataset the feature that tells apart pony and non-pony then? Are you saying that all the other things that do have PPP datasets that are linked in the doc, such as all the TF2 voices, or Carl from Aqua Teen Hunger Force, or HAL 9000 in Clipper's Mega, are in fact ponies? Should I update the pic to include them all?
Not mentioning the Dan Vs. dataset because Dan has been, is, and will always be best pony and I miss him too much to make him part of the discussion.
>>
>>37708338
tl;dr
>>
>>37708338
>And you only need to spend a couple days on /mlp/ to realize that I'm not the only one thinking that way. That right there is WHY this thread exists.
You're damn right about that.
>>
>>37708342
Don't be a pedantic bitch about it. Those obviously aren't pony and that's fine since it's Clipper's personal drive and not strictly a PPP one.
You can post about your non-FiM and even non-pony voice preservation projects. In the same vein, someone sharing a G5 dataset would be acceptable too. Talking about it and *asking* for someone else to do it for you is lazy bullshit that shouldn't be tolerated. At least that's my take. Your opinion might differ and reaching hundred threads later is impossible. "Focus on FiM, feel free to share anything else" is the closest we've had to a universally accepted principle and it's annoying as fuck to see faggots try to undermine even that one single rule.
>>
>>37708386
*reaching a consensus
>>
>>37708386
And yet you seething faggot schizos were reeing even when people ITT organised and got together not to beg for G5 datasets, but to make one. Which we did. People made labels, audio was cleaned, and we even had an attempt at models (too little audio for us, but maybe not for 15, so we will be sending them his way I hope). And every step of the fucking way, you seething cunts waste multiple posts sperging out over it. Shut the fuck up pretending it wasn't like that.
>>
>>37708386
>Talking about it and *asking* for someone else to do it for you is lazy bullshit that shouldn't be tolerated
For the record, I absolutely agree with this. It's completely retarded to come here and ask other people to do something you're to lazy to do, and people who only want free labor should fuck off. But most of the anti-g5 posts aren't "we aren't going to do your dataset for you," they're simply "y-y-you can't talk about that here!" It's retarded to try to censor discussion about other datasets when such a thing has been part of the thread since the beginning.
>>
IT'S TIME!
>>37708443
>>
If the author of 1-1 is here just post what you have. You can post your finished project later and it will be added on after the fact just fucking post your shit you lazy faggot.
>>
>>37708457

NEW THREAD
>>37708443
>>
>>37708386
It's your own goddamned spazzoid anti-G5 ilk that constantly drop the "No G5" post 50 times per PPP thread. Don't try to fucking gaslight and retcon past threads with your bullshit "oH wE cAn AlL gEt AlOnG fInE gUyZ."



Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.