Breeze TTS2
An open-weight text-to-speech (TTS) model with 3.5B parameters, ranked first on the Artificial Analysis speech synthesis leaderboard.
Timeline (last 90 days)
The voice stack consists of four core components: Qwen 3.6 35B (a mixture-of-experts model), Whisper, Breeze TTS2 (3.5B parameters), and Celery VAD.
Breeze TTS2 (3.5B parameters), according to the developer's citation, ranks first on the Artificial Analysis leaderboard, surpassing Gemini and ElevenLabs, and is open-weight.
In the 12GB VRAM configuration, Breeze TTS2 requires about 8.5GB of VRAM (weights plus runtime overhead) and must run entirely on the GPU, while the LLM portion takes about nearly 4GB at a 128K context.
After switching Breeze TTS2 from the full BF16 version to audio.cpp and the Q8 quantized version, no difference in sound quality can be heard, generation speed increases to 2 to 3 times the original, and considerable VRAM space is freed up.
All Facts (4)
The voice stack consists of four core components: Qwen 3.6 35B (a mixture-of-experts model), Whisper, Breeze TTS2 (3.5B parameters), and Celery VAD.
50%UnverifiedBreeze TTS2 (3.5B parameters), according to the developer's citation, ranks first on the Artificial Analysis leaderboard, surpassing Gemini and ElevenLabs, and is open-weight.
50%UnverifiedIn the 12GB VRAM configuration, Breeze TTS2 requires about 8.5GB of VRAM (weights plus runtime overhead) and must run entirely on the GPU, while the LLM portion takes about nearly 4GB at a 128K context.
50%UnverifiedAfter switching Breeze TTS2 from the full BF16 version to audio.cpp and the Q8 quantized version, no difference in sound quality can be heard, generation speed increases to 2 to 3 times the original, and considerable VRAM space is freed up.
50%