this post was submitted on 17 Jul 2023
2 points (75.0% liked)

Oobabooga Text Generation

189 readers
1 users here now

Community for Oobabooga / Pygmalion / TavernAI / AI text generation

Let's rebuild our knowledge base here!

Ooba community is still dark on reddit, so we're starting from scratch. https://www.reddit.com/r/Oobabooga/

Subscribe, engage, post, comment!

Helpful links:

https://github.com/oobabooga/text-generation-webui

https://zoltanai.github.io/character-editor/

https://www.chub.ai/characters

Remember to mark NSFW posts.

Other AI communities to check out:

[email protected]

founded 1 year ago
MODERATORS
 

I just discovered this repo, it looks really useful for creating AI voices

https://github.com/rsxdalv/tts-generation-webui

top 9 comments
sorted by: hot top controversial new old
[–] [email protected] 2 points 1 year ago (1 children)
[–] [email protected] 1 points 1 year ago (2 children)

Tried this yesterday and got it installed and i was able to run it but it took some 20-30 gb and the cloning didn't work. It couldn't access my audiofiles. error 2 iirc. Someone made a video with almost the same probs... -> https://yewtu.be/watch?v=lm6AGTiQ25c

[–] [email protected] 2 points 1 year ago (1 children)

I'm going to try this out, but if it's using that much VRAM I might be out of luck

[–] [email protected] 1 points 1 year ago

if you refer to the 20-30 gb it's disk space i meant. it used more than 30gb though.

[–] [email protected] 2 points 1 year ago (1 children)

Dam, I have this issue:

    File "/home/st/.pyenv/versions/3.10.12/lib/python3.10/json/encoder.py", line 179, in default
      raise TypeError(f'Object of type {o.__class__.__name__} '
  TypeError: Object of type PosixPath is not JSON serializable
  [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip. ERROR: Failed building wheel for pyworld

[–] [email protected] 1 points 1 year ago

what repository? maybe pip install pyworld? i'm having a hard time with fulfilling requirements all the time. (i can't code, i tried...)

[–] [email protected] 1 points 1 year ago
[–] [email protected] 1 points 1 year ago (1 children)

Did you try it? For me Bark isn't usable. It's too robotic and the best results i get with the tts-fast (https://github.com/152334H/tortoise-tts-fast) but i was only able to install it on one computer with W10 and no gpu support. On the other machine i couldn't get the requirements satisfied at all.

Same here: https://git.ecker.tech/mrq/ai-voice-cloning but i'm only on a GTX1080.

[–] [email protected] 2 points 1 year ago

I had issues getting to run, I'll come back to it. I have other ways to generate bark audio. I found bark to be by far the most natural sounding, it just sounds like it was recorded on a pc mic from 1999. Silero, elevenlabs, sounds monotone to me.

I haven't tried Tortoise yet, I'll have to try that!