The story’s score and claims

Comments on No. 59 of 106 · naklecha.com · edition of Thursday 12 November 2026

Caramel: a speech-to-speech conversation model trained only on audio

Summary

Caramel-1 is an open-source 413M-parameter speech-to-speech conversation model trained only on about 11,500 hours of public audio, with no text at all. Its training data was filtered to match the conversation statistics of Dessa's examples, and it generates speech in real time, including on a MacBook Pro.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules