We release a technical blog on training PocketTTS, our 100M parameter on-device TTS, through…
@kyutai_labs28K followers
24,596 views
11 Comments
Comments from the original post on X (Twitter)

@kyutai_labs
PocketTTS speaks in continuous latents, one every 80 ms, and a small head turns noise into the next latent in a single forward pass. We used to train with LSD, another one-step flow matching loss, but this required Jacobian-vector products at training time. This is not needed anymore with the drifting recipe.
11 likes

@kyutai_labs
The key ingredient we came up with to make it work is a learned kernel temperature. The full write-up with the recipe and all the ablations is here : Code: Original Drifting paper:
21 likes

@ainafahy
Can I just say this little guy is so cute
4 likes

@LastResort48
@grok which language it supports?
1 like

@stcisgood
nice work playas

@francarloerra
🙂

@emankhoubian
less than 1% WER on TTS means a speech recognizer heard the words. it says nothing about whether the cloned voice still sounds right at minute 20, which is where synthetic voices break.

@yjelid
@YentiCollin @samsongvch

@ViVL787033
Есть поддержка realtime?

@mdzor
@LocallyAIApp 🙈

@Semonic_
Awesome i love this project I just wish you guys jad used more and better training data