feat(llama-cpp): serve Qwen3-TTS through the llama.cpp backend - #11392

Open

localai-bot wants to merge 18 commits into

feat/llama-cpp-qwen3-tts

Commits on Aug 6, 2026

feat(llama-cpp): implement the TTS and TTSStream RPCs
mudler
committed
cd42e80
fix(llama-cpp): mask non-codec tokens for Qwen3-TTS generation
mudler
committed
5da3b2f
chore(llama-cpp): bump pin to 9de0fcf2b and drop the TTS codec mask
mudler
committed
b04de70
fix(llama-cpp): clear the TTS slot sequence between requests
mudler
committed
1941a8b
feat(llama-cpp): expose max_frames for TTS requests
mudler
committed
60526ff
fix(llama-cpp): send the TTS sample rate up front, and tidy three review items
mudler
committed
873d3c5
build(llama-cpp): let unpatched forks opt out of the TTS task
mudler
committed
1c5ff4e
fix(config): keep a TTS speaker-encoder projector out of vision detection
mudler
committed
e07fc57
feat(gallery): add Qwen3-TTS entries for the llama-cpp backend
mudler
committed
ea1fdc2
docs: cover Qwen3-TTS on the llama-cpp backend
mudler
committed
67f1421
fix(config): register llama-cpp as a TTS and voice-cloning backend
mudler
committed
9805e62
fix(gallery): declare what nemotron-3-nano-omni actually accepts
mudler
committed
bb392fb
fix(importers): import a Qwen3-TTS GGUF repo as TTS, not chat
mudler
committed
aac507d
fix(llama-cpp): stop non-TTS models crashing on the new pin
mudler
committed
f48efa4