Hugging Face's logo
Aayan Mishra's picture
58 12 39

Aayan Mishra

Spestly
horiavarlan's profile picture DevArth's profile picture Triangle104's profile picture
·
https://aayan-mishra.vercel.app/

AI & ML interests

None yet

Recent Activity

published a model about 12 hours ago
Spestly/Kyro-n1.1-3B
reacted to hexgrad's post with 🔥 about 15 hours ago
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.
upvoted an article about 17 hours ago
G2P Shrinks Speech Models
View all activity

Organizations

Stanford AI's profile picture OpenVINO Toolkit's profile picture LLMs's profile picture C4AI Community's profile picture Hugging Face Discord Community's profile picture Data Tonic (Alignment Lab)'s profile picture Odyssey Labs's profile picture Open-Neo's profile picture Lambda Go Labs's profile picture

Collections 4