Chroma 1.0 does not use ASR pipelines, reduces drift, and improves dialogue coherence

Chroma 1.0 released: the world's first open-source end-to-end real-time speech-to-speech model

- Abstract Chroma 1.0 is an end-to-end real-time speech-to-speech model trained by FlashLabs and fully open-sourced, enabling personalized voice cloni...