Top stories

Live feed of AI lab newsrooms. First-party announcements, traced to the source.

Feed Chatter About Privacy Connect RSS

© 2026 forck.live

Top stories

Top stories

Kyutai — forck

The last 8 weeks

3
postsFirst-partypeak 2
0
outlets reportedOn 0 posts

@kyutai_labs·RSS

Lead story

Kyutai· 11 days ago

Diffusable Latents from Structure-Agnostic Distillation

|

|

@kyutai_labs RSS

Announcements Chatter

Kyutai

Announcements Chatter Models The week Connect your AI

Announcements Chatter

CompaniesPerplexity Anthropic Luma Runway Apple Google Research GitHub Kyutai

The last 8 weeks

3
postsFirst-partypeak 2
0
outlets reportedOn 0 posts

Pool, then distill: one vector per image is enough to make a latent diffusable, whatever its shape.

Every post, as it lands

11 days agoBuilding a Pocket TTS with a drifting objective
11 days agoDiffusable Latents from Structure-Agnostic Distillation
25 AugPocket TTS training code released
10 JulMuScriptor: Automatic Multi-instrument Transcription
1 JulSurflo: Consistent 3D surfaces from a global state
18 JunThe FID Lottery
10 JunPost-training speech models for better interactivity
26 MayKairos: Understanding Data Temporality Impact on LLM pre-training
20 MayIntroducing KE:SAI
4 MayPocket TTS now supports six languages
30 AprMoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models
28 AprARC-Encoder: learning compressed text representations for LLMs

Everything else, by date

Building a Pocket TTS with a drifting objective28 Sep
Pocket TTS training code released25 Aug
MuScriptor: Automatic Multi-instrument Transcription10 Jul
Surflo: Consistent 3D surfaces from a global state1 Jul
The FID Lottery18 Jun
Post-training speech models for better interactivity10 Jun
Kairos: Understanding Data Temporality Impact on LLM pre-training26 May
Introducing KE:SAI20 May
Pocket TTS now supports six languages4 May
MoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models30 Apr
ARC-Encoder: learning compressed text representations for LLMs28 Apr
OVIE: One View Is Enough14 Apr

+16 more postsShow fewer

Invincible Voice online demo released24 Feb
Hibiki-Zero: Simultaneous Speech-to-Speech Translation Without Aligned Data12 Feb
Pocket TTS: a high-quality TTS with voice cloning that runs on CPU13 Jan
CASA: The return of the cross-attention23 Dec

Ask your AI

Your brief, straight from the record

Connect once, then ask Claude or ChatGPT what changed for the tools you use. Every item is cited to its record.

+ any MCP app

Connect your AI →

Every post, as it lands

11 days agoBuilding a Pocket TTS with a drifting objective
11 days agoDiffusable Latents from Structure-Agnostic Distillation
25 AugPocket TTS training code released
10 JulMuScriptor: Automatic Multi-instrument Transcription
1 JulSurflo: Consistent 3D surfaces from a global state
18 JunThe FID Lottery
10 JunPost-training speech models for better interactivity
26 MayKairos: Understanding Data Temporality Impact on LLM pre-training
20 MayIntroducing KE:SAI
4 MayPocket TTS now supports six languages
30 AprMoshiRAG: Asynchronous Knowledge Retrieval for Full-Duplex Speech Language Models
28 AprARC-Encoder: learning compressed text representations for LLMs

Ask your AI

Your brief, straight from the record

Connect once, then ask Claude or ChatGPT what changed for the tools you use. Every item is cited to its record.

+ any MCP app

Connect your AI →

Neural audio codecs: how to get audio into LLMs

21 Oct

Kyutai TTS 1.6B3 Jul

Kyutai TTS and Unmute now open-source3 Jul

Kyutai Speech-To-Text released as open-source19 Jun

Unmute: Make LLMs listen and speak22 May

Helium 1: a modular and multilingual LLM30 Apr

MoshiVis: Teaching Moshi to Converse about Images21 Mar

Simultaneous, on-device, high fidelity speech-to-speech translation with Hibiki10 Feb

Announcing Helium-1 Preview13 Jan

Moshi open-source release: run Moshi locally!18 Sep

Meet Moshi, the first real-time voice AI3 Jul

Hello Kyutai!17 Nov