From the source

Pocket TTS training code released

If you have a GPU and some speech data, it's now easy to train your own Pocket TTS model.

Other stories

Building a Pocket TTS with a drifting objective
One-step TTS sampler with drifting objective
Kyutai · 28 SepResearch paper
Diffusable Latents from Structure-Agnostic Distillation
One vector per image makes latents diffusable regardless of shape.
Kyutai · 28 SepResearch paper
MuScriptor: Automatic Multi-instrument Transcription
Kyutai · 10 Jul
Surflo: Consistent 3D surfaces from a global state
Kyutai · 1 Jul

Announced

25 Aug 2026

Detected

09 Oct 2026 13:14 UTC (+44d after)

Kyutai · 18 Jun

Kyutai · 10 Jun