togethercomputer/CoderForge-Preview
Viewer • Updated • 827k • 6.68k • 160
jinwon kim
whooray
AI & ML interests
None yet
Recent Activity
liked a dataset about 1 hour ago
togethercomputer/CoderForge-Preview updated a dataset about 2 months ago
whooray/xpqa published a dataset about 2 months ago
whooray/xpqaOrganizations
liked a
dataset about 1 hour ago
updated a
dataset about 2 months ago
published a
dataset about 2 months ago
liked a
model 3 months ago
liked a
model 6 months ago
Improve language tag
#1 opened 12 months ago
by 
lbourdois
upvoted a paper about 1 year ago
updated a
model about 1 year ago
published a
model about 1 year ago
reactedto benjamin-paine's post with 🔥 about 1 year ago
Post
3120
Zonos is flying up the trending tab, and for good reason - it's the most expressive and emotive open-source TTS I've used to date. I'm happy to say it's now supported in Taproot, with added long-form synthesis support and other goodies.
Try it here: https://huggingface.co/spaces/benjamin-paine/zonos-longform
Getting started with Zonos in Taproot is easy; with a working CUDA toolkit and Python/Pip installation, all you have to do is:
See more on GitHub at https://github.com/painebenjamin/taproot/
Try it here: https://huggingface.co/spaces/benjamin-paine/zonos-longform
Getting started with Zonos in Taproot is easy; with a working CUDA toolkit and Python/Pip installation, all you have to do is:
apt install espeak-ng
pip install taproot
taproot install speech-synthesis:zonos-transformer
taproot invoke speech-synthesis:zonos-transformer --text "Hello, world!"See more on GitHub at https://github.com/painebenjamin/taproot/
- 2 replies
liked a
model about 1 year ago
updated a
model about 1 year ago
published a
model about 1 year ago
reactedto hexgrad's post with 👍 about 1 year ago
Post
6231
I wrote an article about G2P: https://hf.co/blog/hexgrad/g2p
G2P is an underrated piece of small TTS models, like offensive linemen who do a bunch of work and get no credit.
Instead of relying on explicit G2P, larger speech models implicitly learn this task by eating many thousands of hours of audio data. They often use a 500M+ parameter LLM at the front to predict latent audio tokens over a learned codebook, then decode these tokens into audio.
Kokoro instead relies on G2P preprocessing, is 82M parameters, and thus needs less audio to learn. Because of this, we can cherrypick high fidelity audio for training data, and deliver solid speech for those voices. In turn, this excellent audio quality & lack of background noise helps explain why Kokoro is very competitive in single-voice TTS Arenas.
G2P is an underrated piece of small TTS models, like offensive linemen who do a bunch of work and get no credit.
Instead of relying on explicit G2P, larger speech models implicitly learn this task by eating many thousands of hours of audio data. They often use a 500M+ parameter LLM at the front to predict latent audio tokens over a learned codebook, then decode these tokens into audio.
Kokoro instead relies on G2P preprocessing, is 82M parameters, and thus needs less audio to learn. Because of this, we can cherrypick high fidelity audio for training data, and deliver solid speech for those voices. In turn, this excellent audio quality & lack of background noise helps explain why Kokoro is very competitive in single-voice TTS Arenas.
updated a
model over 1 year ago
liked a
dataset over 1 year ago
upvoted an article over 1 year ago
Article
Navigating Korean LLM Research #1: Models
•
26
liked a
dataset over 1 year ago
Create README.md
#1 opened over 1 year ago
by 
rahhal
updated a
dataset over 1 year ago