← Home ← Back to /mlp/

Thread 42320976

293 posts 80 images /mlp/
Anonymous No.42320976 >>42320992 >>42330333 >>42341924 >>42344114 >>42381781
Pony Preservation Project (Thread 155)
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, drawings and text for any person or character using existing audio recordings, artwork and fanfics as a reference. As you can surely imagine, AI pony voices, drawings and text have 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:
docs.google.com/document/d/1PDkSrKKiHzzpUTKzBldZeKngvjeBUjyTtGCOv2GWwa0/edit
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/1y1pfS0LCrwbbvxdn3ZksH25BKaf0LaO13uYppxIQnac/edit
An in-depth repository of tutorials, resources and archives.

>Online speech generation
haysay.ai
alpha.15.dev

>Active tasks:
Research into animation AI
Research into pony image generation

>Latest developments:
pastebin.com/4p00iUZM

>The PoneAI drive, an archive for AI pony voice content:
drive.google.com/drive/folders/1E21zJQWC5XVQWy2mt42bUiJ_XbqTJXCp

>Clipper’s Master Files, the central location for MLP voice data:
mega.nz/folder/jkwimSTa#_xk0VnR30C8Ljsy4RCGSig
mega.nz/folder/gVYUEZrI#6dQHH3P2cFYWm3UkQveHxQ
drive.google.com/drive/folders/1MuM9Nb_LwnVxInIPFNvzD_hv3zOZhpwx

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

Last Thread:
>>42161191
Anonymous No.42320989
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: docs.google.com/document/d/1PDkSrKKiHzzpUTKzBldZeKngvjeBUjyTtGCOv2GWwa0/edit
Main: docs.google.com/document/d/1y1pfS0LCrwbbvxdn3ZksH25BKaf0LaO13uYppxIQnac/edit

>Where can I find the AI text-to-speech tools and how do I use them?
A list of TTS tools: docs.google.com/document/d/1y1pfS0LCrwbbvxdn3ZksH25BKaf0LaO13uYppxIQnac/edit#heading=h.yuhl8zjiwmwq
How to get the best out of them: docs.google.com/document/d/1y1pfS0LCrwbbvxdn3ZksH25BKaf0LaO13uYppxIQnac/edit#heading=h.mnnpknmj1hcy

>Where can I find content made with the voice AI?
In the PoneAI drive: drive.google.com/drive/folders/1E21zJQWC5XVQWy2mt42bUiJ_XbqTJXCp
And the PPP Mega Compilation: docs.google.com/spreadsheets/d/1T2TE3OBs681Vphfas7Jgi5rvugdH6wnXVtUVYiZyJF8/edit

>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.
2020 pony.tube/w/5fUkuT3245pL8ZoWXUnXJ4
2021 pony.tube/w/a5yfTV4Ynq7tRveZH7AA8f
2022 pony.tube/w/mV3xgbdtrXqjoPAwEXZCw5
2023 pony.tube/w/fVZShksjBbu6uT51DtvWWz

>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?
pony.tube/w/mqJyvdgrpbWgZduz2cs1Cm

PPP Redubs:
pony.tube/w/p/aR2dpAFn5KhnqPYiRxFQ97

Stream Premieres:
pony.tube/w/6cKnjJEZSCi3gsvrbATXnC
pony.tube/w/oNeBFMPiQKh93ePqTz1ns8
Anonymous No.42320992 >>42327124 >>42393483
>>42320976 (OP)
Anchor.
Anonymous No.42321069
last thread songs
>>42318746
>https://u.pone.rs/KbiNvzqK.mp3
Solitary Summer Dream by suno user testediserie.
>>42317966
>Beatles' With a Little Help from My Friends with slightly modified lyrics
>https://u.pone.rs/ODLJbBek.flac
>>42310923
>>https://u.pone.rs/NlnRoRSa.mp3
>Ghost singing Past Due - Xenophobia (aka unofficial theme song of Stellaris)
Anonymous No.42321697 >>42321921 >>42322540
What are digital mares doing right now?
Anonymous No.42321921 >>42322337 >>42322540 >>42346150
>>42321697
mares marrying mares?
Anonymous No.42322337 >>42342013
>42318380
>Hi HydrusBeta, Im getting error when using the sovits 4.0 Spitfire model with 'reduce hoarsness' and 'apply nsf_higan' setting, and it works if I turn these two settings off.
>>42321921
mares booping mares
Anonymous No.42322540 >>42322553 >>42322868
>>42321697
>>42321921
Marrying digital stallions and having little digital foals.
Anonymous No.42322553 >>42323811
>>42322540
What types of digital names do they give them?
Anonymous No.42322868 >>42323811 >>42326533
>>42322540
Analog ponies don't get any love these days
Anonymous No.42323811
>>42322868
Only if you want some of that *phat* kick

>>42322553
Actor filter now applies to memory generation (Horses no longer generate memories unless explicitly enabled, for example). - Alpha 16
Anonymous No.42324032 >>42325143
Anonymous No.42324845 >>42325143
https://github.com/Deep-unlearning/Finetune-Dia-TTS?tab=readme-ov-file
Anonymous No.42325143
>>42324845
>multi-turn conversational dataset
uhh, so this is supposedly able make multiple voices using single model? or am I understanding it wrong?
>>42324032
con weekens are always terrible times for getting anything creative done.
Anonymous No.42326428 >>42327934 >>42328836 >>42329567 >>42397505
Bump.
Anonymous No.42326533 >>42326789
>>42322868
I would, but there are no analog ponies in my area.
Anonymous No.42326789 >>42327131
>>42326533
what about the facehoof ads, with the lonely Crystal Kingdom milf mares?
Anonymous No.42327124
>>42320992
Hey, the ai group kyutai has published their TTS model on hugging face (it's the people who made this website https://unmute.sh/)
https://huggingface.co/collections/kyutai/text-to-speech-6866192e7e004ed04fd39e29
Anonymous No.42327131
>>42326789
Could be worth a shot.
Anonymous No.42327934
>>42326428
mare
Anonymous No.42328836
>>42326428
oy
Anonymous No.42329567
>>42326428
sleep bump
Anonymous No.42330160 >>42330929 >>42334201
>9
Deploying bump mare
Anonymous No.42330333 >>42330347 >>42342013
>>42320976 (OP)
I've noticed that the files which you can download from Hay Say regularly cut off the last second or so of the generated audio. I don't know why, but this bug has happened a couple of times already. Has anyone else noticed that too?
Anonymous No.42330347 >>42331030
>>42330333
uhh, strange, I never had that happening to me. Is it happening with all models? does changing file format (mp3/wav) also cuts off audio ?
Is it happening with voice converted audio or TTS?
Anonymous No.42330929
>>42330160
joy of ai amre
Anonymous No.42331030
>>42330347
It happened to me with the controllable talknet, text to speech, regardless of the character choice. The audio output I usually picked was flac.
Anonymous No.42331386 >>42331769
How can I help support heysay. I'm retarded but I have money.
Anonymous No.42331769 >>42332298
>>42331386
redeem
Anonymous No.42332298
>>42331769
Anonymous No.42332940 >>42333632
>nein
Anonymous No.42333632
>>42332940
Almost again.
Anonymous No.42334079 >>42334201 >>42334753 >>42336596
>9
Secondary bump mares initiated
Anonymous No.42334201
>>42334079
>>42330160
Man, they're adorable.
Anonymous No.42334753
>>42334079
bumpmares? kino
Anonymous No.42335216 >>42335721 >>42335724
so it's mare
Anonymous No.42335721
>>42335216
Anonymous No.42335724
>>42335216
Indeed.
Anonymous No.42335807 >>42336334 >>42337053
this is all mare's fault
Anonymous No.42336334 >>42336706
>>42335807
How so?
Anonymous No.42336596
>>42334079
Hmm
Anonymous No.42336706 >>42337053 >>42339264
>>42336334
she knows what she did
Anonymous No.42336973
care for mare
Anonymous No.42337053
>>42335807
>>42336706
They both need a hug.
Anonymous No.42337503 >>42338389 >>42339402
>9
Third bump mare deployed. She blep.
Anonymous No.42337921 >>42338643 >>42339711 >>42342013
Is haysay kill?
Anonymous No.42338389
>>42337503
nine
Anonymous No.42338604
pre sleep bump
Anonymous No.42338643
>>42337921
Currently not responding. Probably just a temporary derp.
Anonymous No.42339264
>>42336706
How many lemons did she eat?
Anonymous No.42339402
>>42337503
Cute blep mare.
Anonymous No.42339711 >>42339743 >>42342013
>>42337921
HydrusBeta, you ok there buddy?
Anonymous No.42339743 >>42342013
>>42339711
https://www.youtube.com/watch?v=T3AYupgiaVs
Anonymous No.42340064 >>42341037
uhoh bump
Anonymous No.42340329
I cast bump spell
Anonymous No.42341037 >>42341478
>>42340064
Indeed.
Anonymous No.42341478 >>42428061
>>42341037
Anonymous No.42341924 >>42349364 >>42351447
>>42320976 (OP)
>>Active tasks:
>Research into animation AI
So what came from the .fla animation exports? I remember someone trying to dump every resource into a series of .svgs and .pngs. What was the goal?
HydrusBeta No.42342013 >>42343816
>>42337921
>>42339711
>>42339743
Sorry for the slow response. Things have been busy lately. Hay Say is back up, now.

>>42322337
The 'reduce hoarseness' option isn't working because Spitfire is a v2 model and apparently the command line arguments were changed and enhanced for that option in v2. Instead of passing "--f0_mean_pooling" to inference_main.py, Hay Say should be passing "--f0_predictor crepe". There are other f0_predictor options too and several other, new command line options in v2 that I could expose in the UI. I'll put this on my to-do list as a bug fix + enhancement.

>>42330333
I haven't encountered this issue myself. Does the audio playback in the UI also cut off the last second too or does that only happen in the downloaded audio file? What length of text input do you use (e.g. is it like a sentence, or more like a couple of paragraphs?)
Anonymous No.42342395 >>42343094 >>42345480
>10
Anonymous No.42342801
>oy
Anonymous No.42343094
>>42342395
Anonymous No.42343278 >>42344753
>thread is hidden by my filters
>I can't find a single part of the OP that actually trips my filters
Anonymous No.42343479
Found another nice sounding sudo song, so I've test out if it sounded nicer in Cadance or Chrysalis vocals.
cadance - [Bee] - by jhkjk
>https://u.pone.rs/VpfpLRXE.mp3
chrysalis - [Bee] - by jhkjk
>https://u.pone.rs/PMZdYYDs.mp3
Anonymous No.42343816
>>42342013
>Does the audio playback in the UI also cut off the last second too
No, it's just in the downloaded version of the files. The on-site audio files play perfectly.
>What length of text input do you use
Usually a paragraph with a couple of sentences. Always somewhere between 30 and 60 seconds of speech output.
Anonymous No.42344114 >>42349575
>>42320976 (OP)
Clipper, or some other archive Anons. I've downloaded the Mega for the episode clips and there seems to be few folders missing.
Season 2:
4, 10, 14, 25, 26
Season 4:
21
Season 5:
6, 9
Season 6:
24
Season 7:
13
Season 8:
5, 6, 7 ,8, 9, 10, 11, 13, 19
Season 9:
22, 23, 24, 25, 26
Anonymous No.42344614
>nein
Anonymous No.42344753
>>42343278
>Not filtering manually
Anonymous No.42345480
>>42342395
Up.
Anonymous No.42345726 >>42346056
Safety bump.
Anonymous No.42346056 >>42346150 >>42346613
>>42345726
Anti homosexual bump
Anonymous No.42346150 >>42346155 >>42357075
>>42321921
>>42346056
Why WOULD mares marry mares?
Anonymous No.42346155
>>42346150
Everyone loves mares, ever mares.
Anonymous No.42346613 >>42347117
>>42346056
Anti page 9 bump.
Anonymous No.42347117 >>42347695 >>42350684
>>42346613
>9 once more
Guess it's time for the fourth bump mare
Anonymous No.42347445
gonna look for some more songs to convert, wish me luck ziggas
Anonymous No.42347695
>>42347117
Thank you for your service.
Anonymous No.42348418 >>42348447 >>42348869 >>42350215 >>42355556
You guys seen this shit? A real-time conversational AI that you can speak to like a person with prompting to define the character and the ability to uploaded voices.
https://unmute.sh/

If we get Twilight Sparkle on this, I will be so happy.
Anonymous No.42348447 >>42348633
>>42348418
>no hooves
Anonymous No.42348633
>>42348447
It can't be that hard to make a pony version out of it, right?
Anonymous No.42348869 >>42350215
>>42348418
They had TTS and voice recognition just published recently on huggingface (in smaller and larger model versions).
https://kyutai.org/next/tts
https://kyutai.org/next/stt
I haven't seen anybody doing some easy webui for it yet.
Anonymous No.42349364 >>42349395 >>42349575
No one?
>>42341924
So that's it? The anons that were involved in this have moved on?
Anonymous No.42349395 >>42349556 >>42350378
>>42349364
Yes. And if anyone tries to say otherwise, you can look through the past year of threads yourself. The most that's happened in recent times is a few song covers, and some folks bringing up AI models that no one takes action to test. Even 15 returning only revitalized this thread for maybe ~24 hours.
Welcome to the bump general.
Anonymous No.42349556 >>42349926 >>42350378 >>42350965 >>42353758
>>42349395
As someone who used do content in early threads, Im currently limited by what I can do with my 5+ year old pc, adding to that the absolute lack of spare time, the result is what it is.
I would image the fact that even when somebody does make something here, there is 0 response to give people their (you)s, even in /create/ and /bale/ there will be one response of "hey, that was nice work Anon", is bit demotivational.
Clipper No.42349575 >>42350215
>>42344114
Those episodes are covered by the studio leaks in the special source folder. The audio from those is theoretically perfect so no real need for duplicates.

>>42349364
I'm still here, just a lack of things I specifically can do aside from maintaining the data resources. Mare Fair prep eats most of my time these days.
There's certainly still value in this thread still existing as a repository if nothing else.
Anonymous No.42349926 >>42350215 >>42350378
>>42349556
Totally understandable, and I don't blame anyone for the thread's current state. I think this general at its peak was a bit of a lightning in a bottle situation of its own. A hell of a lot of cool pony stuff came out of this though.
Anonymous No.42350215 >>42352227
>>42349575
>in the special source folder
Thanks. I am just missing S2E04, S8E19, S8E23, S9E22, S9E24. I Don't really care for s8 and s9 episodes clips other than just to have it for completion, but I would like to have audio of Luna first proper appearance.
>>42349926
Yeah, life is life. One thing that is positive there are still random projects happening outside the thread like >>42348869 >>42348418 new TTS model.
Anonymous No.42350378
>>42349395
>>42349556
>>42349926
Non-pony parasocial reoccurrences are orthogonal to artistic intent. Why would we even "think" that the push-pull *techplot* control drama is cutemark inclined? Please consider the pity.

The bump is the implicit cost of anon's business's flank.
Anonymous No.42350684
>>42347117
ehh
Anonymous No.42350965
>>42349556
>there is 0 response to give people their (you)s
That's a problem across the board, unfortunately. I've seen that a lot recently.
Synthbot No.42351447 >>42351937 >>42357075 >>42358220
>>42341924
Dumping all of the SVGs/PNGs would have taken up way too much space, so we instead have:
- An XFL version of all the FLAs.
- A tool to dump any still or sequence from any frame in an XFL.

>xfl files
https://drive.google.com/drive/folders/1kk8Xb5Xht4wahyHYOIVpMRtB69eg3Yhl?usp=drive_link
>tool to dump images from XFL files
https://drive.google.com/drive/folders/1xROohg-r_10asrz3xGhh7iMSXIa0vyQX?usp=drive_link
>source code for the tool
https://github.com/synthbot-anon/xflsvg

If you search desu for posts from me containing "xflsvg", you'll see some example commands for the tool.
https://desuarchive.org/mlp/search/text/xflsvg/username/Synthbot/
>python3 -m xflsvg INPUT_XFL_FOLDER/.xfl OUTPUT_FOLDER/.svg --use-document-attrs
>python3 -m xflsvg MLP506_439/.xfl outputs/.svg --use-document-attrs
>python3 -m xflsvg MLP506_439/.XFL pieces/.samples
>python3 -m xflsvg MLP506_439/.XFL pieces/.samples --render-sample-shapes
>python3 -m xflsvg animation-assets-sep-01-2022/.xfl --batch CleanAnimationsFolder/.gif --focus mlp-animation-symbol-labels/retain/Characters/Clean/.samples --skip-leading-blanks --no-stills --background '#d6daf0ff'
>https://desuarchive.org/mlp/thread/39220092/#39264940

Maybe I should make an MCP server for this thing so people don't need to learn the complex flags...
Anonymous No.42351937
>>42351447
>Maybe I should make an MCP server for this thing so people don't need to learn the complex flags
Might revitalize interest if people can now play with animations.
Anonymous No.42352215 >>42352696
Up.
Clipper No.42352227
>>42350215
>S2E04, S8E19, S8E23, S9E22, S9E24
Thanks for pointing that out - not sure how those went missing, but they're back now. s2e4 is in the FiM folder, all others in special source.
Anonymous No.42352696
>>42352215
Anonymous No.42352765 >>42353379
>board is getting spammed again
fucking jannies and mods
Anonymous No.42353379
>>42352765
Anonymous No.42353509 >>42409072
https://u.pone.rs/VLjdTEjy.mp3
During the mlpcon there was a panel "Fucking Around Interdenominational" related to meditation and stuff, I wasnt fan of the quality of the guys mic so I've run it by voice converter... and it's still kind of sucky, I tried to swap the outputs that didnt sound like spoken words with TTS clips but now Luna is randomly swapping pitch and tone. I will be working on version 2 sometime in the future but here the version 1 tso there is at least something to listing to.
Anonymous No.42353758 >>42353782
>>42349556
Comes with the territory of people fostering the cultural mindset of "saying positive things = creating an eceleb CIRCLEJERK and that's BAD and WRONG and FURRYDISCORDTWITTEROFFBOARDREEEE" because nobody tells said shitposting retards to fuck off strongly enough.
Anonymous No.42353782
>>42353758
yes, but alas, addressing it will change nothing. a chain is only as strong as its weakest link, and it only takes one shitposter to create bait or samefag and derail a thread for "teh lulz". unfortunately this board primarily consists of weak links.
Anonymous No.42354300 >>42354756
>42353758 >42353782
It's sure summer in here .
Anonymous No.42354756
>>42354300
>too scared to actually link posts he's replying to
Precisely what I'm talking about. Shitposting retards who read something on kym or r/4chan once and thought natural 4chan behavior was nothing more than shitting on everyone and everything and seeing how many slurs you can shart out in a single post.
Anonymous No.42355556 >>42355941
>>42348418
For a moment I thought this would be used by the site to represent Twilight.
Anonymous No.42355941 >>42356665 >>42358574
>>42355556
>used by the site
What do you mean by that? Like /chug/ ai character card?
Anonymous No.42356665
>>42355941
Anonymous No.42357075 >>42360400
>>42351447
Is this something that could be picked up by the PPP VN over in /CHAG/?

So,
>42346056 >42353758
>there is no philia to be found,
>42354756
>no personality cult to emulate against,
>42353782
>and no enumerable community for Anon.

When you greet a being from another world, should you not green them and deny their true name?

>>42346150
But how COULD mares marry mares?

Musubi enclosure required. Getting real tired of this *meta*.
Anonymous No.42357085 >>42358072
>12k dollars separating Anons from being able to run a simulated Equestria
Also, holy shit, used acceleration cards from decades ago still sell on the same level as if they were brand new, this is such bullshit.
Anonymous No.42357574
>nein
Anonymous No.42357832
>pre sleep anti spam bump
Anonymous No.42358072
>>42357085
They cost that much because people keep paying that much for them. I'm sure they could be sold for at least half that and still turn a profit but then nvidia wouldn't be a trillion dollar company now would it? blame crypto, blame the mainstream AI revolution, and blame AMD for being completely fucking dense.
Anonymous No.42358101 >>42358603
https://huggingface.co/nvidia/audio-flamingo-3
Anonymous No.42358220 >>42360400
>>42351447
yea
Anonymous No.42358574
>>42355941
Like an AI generated and perhaps even animated artwork made by the AI site.
Anonymous No.42358603
>>42358101
That's interesting, if I could run it I could see the use of it by dumping the entire library of unnamed audio effects and have proper list of descriptions for them
>https://audioflamingo3.github.io/static/voice/alpaca_eval_3.mp3
lamo at the random twi clip
Anonymous No.42359133
loving ai mares
Anonymous No.42359628
mares
Synthbot No.42360400 >>42361045
>>42358220
On Windows, you need to be running Docker Desktop first and make sure you have WSL installed. I'd guess that you already have everything installed, but you haven't started Docker Desktop. I don't think it starts by default on boot.

>>42357075
Yes, it look relevant for that. If you run xflsvg with an output file that ends with "/.samples", it will show you all of the exportable assets. If the output file ends with "/.svg" or "/.xfl", it'll dump an SVG or XFL file with the asset. In either case, it can dump every frame of that asset, so you can get every pose in an XFL file, and we have the XFL files for several episodes. That could be used for characters, props, and backgrounds.
Anonymous No.42360461 >>42360876
https://mistral.ai/news/voxtral
Anonymous No.42360876
>>42360461
>Natively multilingual speech recognition
That's pretty neat.
Anonymous No.42361045 >>42361050 >>42364554
>>42360400
Oh, I had to run Docker first. It seems I can export whole scenes but trying to focus on symbols doesn't work.
Am I doing this right?
>To specify a timeline, you can append the symbol name in brackets ("/file.xfl[~Octavia*Character]").
Anonymous No.42361050 >>42364554
>>42361045
wrong pic
Anonymous No.42361394 >>42361403
A more American-South accurate farmer Applejack:
https://voca.ro/1bd7rZ0VOPFm
Anonymous No.42361403 >>42362240
>>42361394
>can't understand shit
it's perfect
now make her talk like boomhauer
Anonymous No.42362240
>>42361403
Kek, I've just had the surreal experience of realizing that without using my headphones, this goes from perfectly understandable to utterly impossible to follow for me.
Anonymous No.42362365 >>42363233
How can I train my own models for haysay?
I have a local haysay install and want to try training some voices that aren't there.
Anonymous No.42363047 >>42365382
Bump for the bump general.
Anonymous No.42363233 >>42364106
>>42362365
There should be a model list json file, that lists all current models + their types, I would imagine it should be as simple as training the model and adding it to the list.
One thing I would like to know is how to train the gpt-sovits character mood Trait models, since thats not included in vul training github.
Anonymous No.42364106
>>42363233
Mare
Synthbot No.42364554 >>42365152
>>42361045
>>42361050
Honestly I forgot and can check a bit later. But I think it should be [.Octavia.Character.regex] or [~Octavia*Character.asset].
Anonymous No.42365152 >>42366045
>>42364554
Okay, I just needed to append .asset to the end.
Anonymous No.42365382
>>42363047
Thank you for your service.
Anonymous No.42365409 >>42365468
I get this is the general for the voice stuff, but is anyone aware of any LoRAs for regular generative AI for various characters?
Anonymous No.42365468 >>42365524
>>42365409
For image gen: >>42353451
For text gen: >>42363249
Anonymous No.42365506 >>42365508 >>42365537 >>42365889 >>42366173
Do you think the live action movie will have lines worth extracting for the datasets?
>pic unrelated
Anonymous No.42365508
>>42365506
no, because I don't want to hear new VAs.
Anonymous No.42365524
>>42365468
Thank you Anon.
Anonymous No.42365537 >>42365543 >>42365889
>>42365506
>live action
Please be trolling. Please, PLEASE be trolling.
Anonymous No.42365543
>>42365537
ah, you hide threads too?
Anonymous No.42365889 >>42372212
>>42365506
>>42365537
>live action
Do not want, REEEEE!!!!
Anonymous No.42366045
>>42365152
She's such a pretty mare
Anonymous No.42366173 >>42366640 >>42367119 >>42367247
>>42365506
>Live action horses
Now that would be a stable production.
Anonymous No.42366640
>>42366173
Anonymous No.42367119
>>42366173
Anonymous No.42367247
>>42366173
Not if Hasbro takes the reins.
Anonymous No.42367670
mamre
Anonymous No.42368249
mare mare
Anonymous No.42368799 >>42368812
Just found this:
https://twibooru.org/3466607
Was this already posted here by any chance?
Anonymous No.42368812 >>42368832
>>42368799
pretty sure Ive heard the song before but I dont remember the animation.
Anonymous No.42368832 >>42376717
>>42368812
This is the original:
https://www.youtube.com/watch?v=vsPsbgFlNwk
Anonymous No.42369207 >>42370241
https://huggingface.co/datasets/deepvk/NonverbalTTS
Anonymous No.42369871
>ooo
Anonymous No.42370241
>>42369207
that's a nice dataset, maybe in the future when new type of emotion control is introduced we could use it for ponies
Anonymous No.42370696
one for ai mares
Anonymous No.42371111
>mare mare
Anonymous No.42371795
>9 again
Anonymous No.42372101 >>42372274 >>42372994 >>42389208
I can't believe this STILL didn't get voice acted by Twilight Sparkle.

vocaroo.com/1oHodh5vYKpR

A cutie mark is far more than a mere symbol or identifier, it is the distilled essence of a pony’s very being. It is not simply a reflection of a talent or hobby, nor a role assigned by society. Instead, it stands as an intricate, immutable emblem of individuality, representing a pony’s soul, heritage, and identity in a way that is both deeply personal and profoundly abstract.

This mark is a tapestry of meaning, weaving together culture, ancestry, character, and spirit. It is a flag of individuality, a coat of arms that each pony bears proudly. Like a fingerprint unique to the self, it cannot be replicated or erased. In its permanence, as confirmed in Call of the Cutie, the cutie mark becomes a lifelong affirmation of one’s unique narrative, a sacred banner of identity and self-discovery.

To reduce such a profound symbol to a mere vocational label, as some later depictions in Cutie Pox or Magical Mystery Cure attempt, is to strip it of its true magnificence. A cutie mark is not a job or an obligation; it is a timeless reflection of the harmony between body, mind, and soul. To trivialize its meaning is to misunderstand its transcendent role in expressing individuality and purpose.

Scientifically, one might liken it to a unique genetic code, an expression of existence so layered and intricate that it defies reductive interpretation. Emotionally, it is a beacon, a radiant testament to the miracle of identity and the wonder of self-expression.

Let the cutie mark remain untouched, its beauty unblemished and its meaning untarnished. To honor the cutie mark is to honor the sacred, irreplaceable essence of the individual. It is a celebration of the complexities that define us, a crystallized symbol of the infinite beauty of the soul. Let it forever stand as the brilliant coat of arms it was always meant to be a shining flag of the heart, unfurled in the winds of life. A true snowflake essence known colloquially as snowpity.
Anonymous No.42372212
>>42365889
Wait, that wasn't a joke?
Anonymous No.42372274
>>42372101
that would be an wholesome idea.
Anonymous No.42372994 >>42379998
>>42372101
>I can't believe this STILL didn't get voice acted by Twilight Sparkle.
More people would take vogelfag seriously if he wasn't so egotistical.
Anonymous No.42374187 >>42374820 >>42375367
Hey HydrusBeta, where can I find all the RVC models that are available on HaySay?
(specifically, I'm looking for the Rainbow Dash(s1) model)
I found this https://huggingface.co/hydrusbeta/hay_say_reuploaded_models but it looks like it's missing some of the models that are currently on HaySay. Is there another place where I can find more of the models? or could I bother you to update the repo?
Anonymous No.42374820 >>42375367
>>42374187
+1 to this
HydrusBeta No.42375367 >>42375543 >>42375886 >>42377063
>>42374187
>>42374820
Hello. The Rainbow Dash (s1) model can be found here:
https://huggingface.co/therealvul/RVCv2/tree/main/RainbowDashS1

Direct download links for the .pth and .index files for all the RVC models available in Hay Say are listed in a JSON file in the main repo:
https://github.com/hydrusbeta/hay_say_ui/blob/main/architectures/rvc/character_models.json
There are similar JSON files for the other architectures, too.

Whenever I notice that a file is no longer available (e.g. somebody's MEGA account disappears), I'll re-upload it to the hay_say_reuploaded_models repo. I should go through all of them again sometime; it's been a while since I've checked for broken links.
Anonymous No.42375543
>>42375367
thanks pal
Anonymous No.42375886
>>42375367
thanks man!!
Anonymous No.42376426 >>42377063 >>42380406
https://ecency.com/actor/@blaffy/voice-actors-demand-regulation-on-ai-voice-cloning
>tfw you get 20 to life because you generated Princess Celestia saying she wants your hot monkey dick
Anonymous No.42376717
>>42368832
That's a weird one to be honest. It doesn't really click.
Anonymous No.42377063 >>42380126
>>42375367
Sometimes when I launch docker to use my local haysay instance, I have to redownload the AI voices I was sure I already had on my drive, because I just used them the day before.
How can I make sure those permanently stay on my drive in case >>42376426 makes me have to hunt for this stuff? I'm not seeing a way to just download everything from huggingbox (unless I have to sign in?)
People are really flipping out over this a lot. I just want to generate mare voices. Is that really too much to ask?
Anonymous No.42377443
Up.
Anonymous No.42377841 >>42378477 >>42378926 >>42379562
>early upsies
Anonymous No.42378477
>>42377841
Anonymous No.42378926 >>42379919
>>42377841
Anonymous No.42379562
>>42377841
Anonymous No.42379919
>>42378926
Anonymous No.42379998 >>42380841
>>42372994
More people would take /mlp/fags seriously if you weren't so egotistical.
HydrusBeta No.42380126
>>42377063
Docker *should* be retaining the voices you already downloaded. They are saved in a persistent docker volume named "models". If you have Docker Desktop, you can browse through the files in that volume (pic related). If a voice seems to disappear again, could you check whether its files are still present in the volume? If so, it could indicate that Hay Say is having trouble seeing the files for some reason. If they are actually deleted out of the volume, I'd be a bit surprised.
Anonymous No.42380406
>>42376426
I doubt they will succeed. Corpos will bank on ai to cut costs and effort and they won't let a bunch of voice actors stop them.
Anonymous No.42380841 >>42381348
>>42379998
Anonymous No.42381348 >>42382660
>>42380841
Anonymous No.42381781
>>42320976 (OP)
Anonymous No.42382277 >>42384021
>https://x.com/de5imulate/status/1947024682118488116
Not pony, just some neat looking ai video going around /v/. Made me wish there was a project that could mesh the Elden Ring with the Gen1 DnD like aesthetics.
Anonymous No.42382660 >>42383185
>>42381348
Anonymous No.42383185 >>42384742 >>42385749
>>42382660
Anonymous No.42383561
mare
Anonymous No.42383824 >>42383931 >>42384156
>you can pet the virtual mares
the future cant get here soon enough
Anonymous No.42383931
>>42383824
Anonymous No.42384021 >>42390151
>>42382277
I love that verticality with those clouds and mist. Reminds me of Albert Bierstadt's paintings.
Anonymous No.42384156
>>42383824
>Japan
If that stuff finally reaches my spheres (if it does at all), several decades will have passed. I''m sure of it.
Anonymous No.42384742
>>42383185
Anonymous No.42385314
>bump down the road
Anonymous No.42385749 >>42386634
>>42383185
Anonymous No.42386634 >>42387034 >>42387469
>>42385749
Anonymous No.42387034
>>42386634
mare
Anonymous No.42387469 >>42387881
>>42386634
Anonymous No.42387881
>>42387469
Anonymous No.42388282
Need ai mares
Anonymous No.42388536 >>42388552 >>42388768 >>42388833 >>42392910 >>42398454 >>42404456
Is PPP behind the HLVR series, or is that from someplace else?

I was kinda sad there was no HLVR the last con. There's so much cool potential. Left 4 Dead, but with anonfillers instead of zombies. Or maybe Portal, where Anon has to lead mares through the puzzles.
Anonymous No.42388552 >>42388768
>>42388536
No, we do nothing but bump.
Anonymous No.42388768
>>42388552
im doing stuff, just figured out how get offline gpt-sovits to work. Fun fact, Luna model cannot for the life of hers say the word "sit", so a that word need to be substituted with "seat" instead.
>>42388536
I would love to use real life rvc voice changer, trouble is none of the programs "swap this mic input with input from this program" work on my piece of shit pc.
Anonymous No.42388833
>>42388536
It worked on some of the tech, but it's the gmares thread that's behind it.
Anonymous No.42388894
>working on an asmr kind of thing
>randomly get this error in second part of sentence
>https://voca.ro/1jNjKreVJjrS
huh, weird.
Anonymous No.42389208
>>42372101
So is someone going to do this or not?
Anonymous No.42389386 >>42389819 >>42389819 >>42390580 >>42391024 >>42391549 >>42394085
Precautionary late night bump.
Anonymous No.42389819
>>42389386
>>42389386
Anonymous No.42390151 >>42391168
>>42384021
https://files.catbox.moe/wzediw.mp4
"Blow (on) Anon and become his Waifu for life." Chat is this true?
Anonymous No.42390580
>>42389386
Anonymous No.42391024
>>42389386
Anonymous No.42391168 >>42392024
>>42390151
>when your waifu is inspired by the perfectly white teeth and just teleports them out of you mouth to make a necklace.
Anonymous No.42391549
>>42389386
mared
Anonymous No.42391993
>9
fr fr
Anonymous No.42392024
>>42391168
https://www.youtube.com/watch?v=Ns2dyze1yu8
Anonymous No.42392529
its mare
Anonymous No.42392910 >>42398999
>>42388536
What's HLVR?
Anonymous No.42392965
https://www.boson.ai/technologies/voice
https://github.com/boson-ai/higgs-audio
Anonymous No.42393483 >>42393621 >>42393678 >>42394168
>>42320992
The button's mom dataset is finally complete:
https://mega.nz/file/uvhSXRRY#1E68yKQEAt4RpzNXt0lwo_2p26ynVqpxfdmXkhY7Mn8
46 minutes of audio total.
Anonymous No.42393621
>>42393483
just found a duplicate audio in the dataset, might have to clean it up and post again. I added in songs that I had to post process at the end of the dataset.
Anonymous No.42393678
>>42393483
>Reupload
https://mega.nz/file/S2ZDTRID#-i5Mqx_vIj0sRgr70W1PJ2HeHMTs4SCR3J_EkUpziOI
Anonymous No.42394085 >>42394797
>>42389386
Anonymous No.42394168
>>42393483
Nice work Anon
Anonymous No.42394409
>quick bump
Anonymous No.42394797 >>42395188
>>42394085
Anonymous No.42395188
>>42394797
mare
Anonymous No.42395647
amre
Anonymous No.42396264
nein
Anonymous No.42396664
eepy bump
Anonymous No.42397505
>>42326428
Anonymous No.42397878 >>42399739
Up.
Anonymous No.42398161
too much summer posting
Anonymous No.42398454 >>42398999
>>42388536
>HLVR
Half Life VR?
Anonymous No.42398702
saturday mare
Anonymous No.42398999
>>42392910
>>42398454
>https://pony.tube/w/nA7nm44BorPe2U3XUnUczY
There's more, but that one was the first. It's Half-life in VR, with "AI ponies"
Anonymous No.42399739 >>42401131
>>42397878
Anonymous No.42400166
>pre eepy bump
Anonymous No.42401131 >>42401471
>>42399739
Anonymous No.42401471 >>42401884
>>42401131
Anonymous No.42401884 >>42402424
>>42401471
Anonymous No.42402424 >>42403069
>>42401884
Anonymous No.42403069 >>42403600
>>42402424
Anonymous No.42403600 >>42403607
>>42403069
Anonymous No.42403607 >>42406534
>>42403600
Anonymous No.42404456
>>42388536
I'm the one that runs that, couldn't do anything this last /mlp/ con because the date didn't work for me and I was getting a little burnt out, but I agree there's still a lot of potential there
we'll keep making them for as long as we're having fun and we still see the potential in it
kind of glad to hear someone missed it, lol. Don't worry, we'll be back!
Anonymous No.42404580 >>42404755
Fascinating thread.
Anonymous No.42404755 >>42405235 >>42409072
>>42404580
AI Voice working is slow, even when using reference audio only one out of 15-30 clips has the correct tone to be used.
Anonymous No.42405040 >>42405386
>>42390593
Oh cool, Anon from /chag/ finished his Visual Novel ai chat game/plugin.
Anonymous No.42405235
>>42404755
I know, I know. It's just that the thread speed makes it look like a ghost town.
Anonymous No.42405386 >>42405978
>>42405040
I wonder if voices could be easily added to it.
Anonymous No.42405978
>>42405386
I would imagine some kind of API would need to be modded into it, to pass along info on what pony is currently talking on the screen along with the text, than it would need to be connected to some light TTS model is able to quickly swap between character voice models.
So technically yes, but it would need somebody more advanced figuring it out to make it work.
Anonymous No.42406534
>>42403607
Anonymous No.42407637
Bumperino
Anonymous No.42409072
>>42353509
>>42404755
>https://files.catbox.moe/bbmjgz.mp3
Alright, complete re-dub using the Luna tts, so it's all in the same tone all throughout, Ive even added some nice meditation bells in the background to help chill out into it.
Anonymous No.42409711 >>42410461
eepy bump
Anonymous No.42409936 >>42412846
TTS-1 Technical Report
https://arxiv.org/abs/2507.21138
>We introduce Inworld TTS-1, a set of two Transformer-based autoregressive text-to-speech (TTS) models. Our largest model, TTS-1-Max, has 8.8B parameters and is designed for utmost quality and expressiveness in demanding applications. TTS-1 is our most efficient model, with 1.6B parameters, built for real-time speech synthesis and on-device use cases. By scaling train-time compute and applying a sequential process of pre-training, fine-tuning, and RL-alignment of the speech-language model (SpeechLM) component, both models achieve state-of-the-art performance on a variety of benchmarks, demonstrating exceptional quality relying purely on in-context learning of the speaker's voice. Inworld TTS-1 and TTS-1-Max can generate high-resolution 48 kHz speech with low latency, and support 11 languages with fine-grained emotional control and non-verbal vocalizations through audio markups. We additionally open-source our training and modeling code under an MIT license.
https://github.com/inworld-ai/tts
Only the training and modeling code. No models
https://inworld-ai.github.io/tts
Examples
The voice cloning sounds pretty good. But it's a product (https://inworld.ai/tts)
>Inworld-TTS-1 Inworld-TTS-1-max
>$5/1M characters $10/1M characters
1 million hours for the pretrain dataset and 200k hours for the SFT dataset
Anonymous No.42410461 >>42410984
>>42409711
early bump
Anonymous No.42410984 >>42411684
>>42410461
so be it
Anonymous No.42411684 >>42412295
>>42410984
Anonymous No.42412295 >>42412616
>>42411684
Anonymous No.42412616 >>42413240
>>42412295
Anonymous No.42412846
>>42409936
>https://inworld-ai.github.io/tts
yeah, the quality of examples is pretty good, I just wish that there would be some companies that instead of going to try making as large possible audio model had instead try to make a smallest yet still good quality tts.
Anonymous No.42413240
>>42412616
Anonymous No.42413714 >>42414217 >>42416276 >>42426955
has anyone tried to vibe coding with ai mares?
Anonymous No.42414217
>>42413714
Anonymous No.42416064 >>42416770
bump
Anonymous No.42416276 >>42420231
>>42413714
Not as far as I know.
Anonymous No.42416770 >>42419785
>>42416064
Anonymous No.42417801 >>42417853 >>42418282
https://x.com/jiqizhixin/status/1951195402096746798
https://arxiv.org/abs/2506.21619
Anonymous No.42417853
>>42417801
https://huggingface.co/IndexTeam
not up yet but will presumably be posted here
Anonymous No.42418282 >>42418376 >>42418828
>https://www.youtube.com/watch?v=OiXO1_Nw_lA
>Yu-Gi-Oh! Master Duel is getting AI live commentary
Just spot this one on /v/, while I am almost certain that the above is going to be totally ass, I kind of like idea of ai pony vtuber to play games with (either, as a back seat commentator or a co-op player 2).
>>42417801
Interesting, but I am worried it once again will be a model requiring some silly amount of 32GB VRAM.
Anonymous No.42418376
>>42418282
Fuck 32GB VRAM would still be doable with two cards. The problem is that we're already at the point where 128GB systems like DGX Spark are bandwidth constrained. I've been so tempted to buy one of those dedicated AMD AI machines.
Anonymous No.42418828 >>42419053
>>42418282
>pic
Is that a Loituma reference?
Anonymous No.42419053
>>42418828
>Not recognising Miku
I know it's a bait, screw you for making me give (you)s
Anonymous No.42419785 >>42419969
>>42416770
Anonymous No.42419969 >>42420496
>>42419785
Anonymous No.42420231
>>42416276
that's a shame. I remember from few months ago someone made a video of "Flutershy makes edutainment podcast about animals", and I felt like something along this line should be more popular.
Anonymous No.42420496 >>42420841
>>42419969
Anonymous No.42420841 >>42422690 >>42423761
>>42420496
Anonymous No.42421183
woow
Anonymous No.42421732
>neign
Anonymous No.42422690
>>42420841
10
Anonymous No.42423523 >>42424084
Where are the mares?
Anonymous No.42423761 >>42425867
>>42420841
Anonymous No.42424084
>>42423523
Sleeping.
Anonymous No.42424639
bumpo
Anonymous No.42425244 >>42425521
I've been trying to use HaySay to get Sunset to voice this but it's not really working out. Anyone who's more experienced with it able to try?
https://files.catbox.moe/cq8g7a.mp3
Anonymous No.42425521
>>42425244
first, tehre is lots of reverb, so I would say to first run it by some kind of de-reverb software (there are few models to download with Ultimate Vocal Remover), than I would say, chop it into 10s clips and run it by RVC (mess around with the pitch, usually going 12 up~down should help but in some occasion I needed to change it by 6th and even 3rd pitch).
If it still not sounds like Sunset, I would suggest to experiment by running one clip by some other character -> than use that output for Sunset model (this will degrade the quality of the audio to some degree, but there is chance it will not be that noticeable).
Alternatively, you can try to use the TTS part of gpt-sovits to redo the audio from scratch.
Anonymous No.42425867
>>42423761
Anonymous No.42425870 >>42426384
>https://huggingface.co/mradermacher/GPT4chan-24B-GGUF
>there are threads on /g/ and /pol/ that are straight up run purely by bot shitposts
strange times we are living in
Anonymous No.42426384 >>42426698
>>42425870
Yeah, the botting has become more and more obvious in the recent weeks. It has hit completely new levels.
Anonymous No.42426698 >>42427107
>>42426384
>https://suptg.thisisnotatrueending.com/fuckai.html
hmm, looks like things are not just bad on posting level but also on the archival websites as well, the LLMs are going bit schizo and try to scrap any and all data from websites even if the webpages do not exist.
Anonymous No.42426950
>nein
Anonymous No.42426955
>>42413714
Kek gib derpy hooves coding companion
Anonymous No.42427107 >>42427170
>>42426698
I'm starting to think this will cause some severe issues down the line.
Anonymous No.42427170
>>42427107
seeing that there are people from Amazon and Google are stating that they don't have enought training data despite having access to pretty much an entire back up of the whole internet, makes me worry how the future text models will get more schizoid from being trained on artificial datasets made by other LLMs.
Anonymous No.42427539
emergency bump
Anonymous No.42428061
>>42341478