AI Briefingllama.cpp Integrates Qwen3-TTS
Key point
The llama.cpp mainline now supports voice cloning with Qwen3-TTS 1.7B.
Details
The Qwen3-TTS-12Hz-1.7B-Base voice cloning implementation has been merged into the llama.cpp mainline.
You can now generate speech using approximately 3 seconds of WAV·MP3 audio as a speaker reference file in the llama-tts binary. Supported languages include:
- English, Chinese, German, Italian
- Spanish, French, Portuguese, Russian
- Japanese, Korean
Current support is limited to the 1.7B Base model, and CustomVoice and VoiceDesign are not supported. The /tts server endpoint is still in draft stage, and the existing llama-tts binary includes breaking changes.
Comparisons of voice similarity, speed, and memory usage against qwen3-tts.cpp or audio.cpp have not yet been published.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.
Related
Halo NeuroAI Releases Sopro V2 Turbo, a 120M-Parameter Open-Source Voice Cloning TTS with Native European Portuguese Support