caramel: open-source speech-only conversational model sounds human but can't hold a coherent chat yet
realmrfakename · x · 2026-11-11
Developer naklecha introduced caramel, an attempt to train a Dessa-style conversational model purely on speech. It sounds very human but can't yet hold a coherent conversation. The model weights and training recipe are open-sourced, with samples of caramel talking to itself released as demos.
More from Models
- Dev claims ~700 tokens/s running Qwen 3 27B on a single RTX 5090 — IgorCarron · 2026-11-11
- Qwen4 Max Preview appears sans benchmarks, rumors call it disappointing — kimmonismus · 2026-11-11
- Inkling 2 Small goes free on OpenCode: 594B MoE, 1M context, multimodal — ZhongRuiqi · 2026-11-11
- Gemini free tier users report usage limit quietly nerfed from ~2% to 6% per prompt — ItsMichaelRay · 2026-11-11
- Researcher lists three explanations for "model nerfing," one intentional — gleech · 2026-11-11
- GPT-4-turbo outputs shorter completions when told it's December, N=477 test — gleech · 2026-11-11