Speech generation · Opt-in
Breeze-TTS-2
A 3B model run by audio.cpp: English and Chinese cloning and voice design, with no Python.
- BreezeBlueBy
- 2Languages
- 24 kHzOutput
- ~4.73 GBDownload
Can do
- Voice cloningYes. YesSpeaks in a voice copied from a short reference clip.
- Voice designYes. YesBuilds a new voice from a written description.
- Emotion controlNo. NoTakes graded emotion from a clip, a vector or text.
- Preset voicesNo. NoShips fixed voices; no reference clip needed.
Languages
2languages
- Chinese
zh - English
en
Runs on
audio.cpp server
isolatedaudio.cppCUDAVulkanApple SiliconCPUROCm · buildPrebuilts: Windows CPU, Vulkan and CUDA; Linux CPU and Vulkan (CUDA or ROCm from a self-build); macOS arm64 Metal.
Engine doc
The Q8_0 GGUF is about 4.73 GiB; allow about 6 GB of dedicated VRAM.
Hardware
| Engine | CUDA | ROCm | Intel Arc | Apple Silicon | Vulkan | NPU | CPU |
|---|---|---|---|---|---|---|---|
| audio.cpp server | Supported | Needs a build | Not supported | Supported | Supported | Not supported | Supported |
Supported Needs a build Experimental
Licence
- Code
- Apache-2.0 (audio.cpp code)
- Weights
- BreezeBlue Research and Non-Commercial License: research and non-commercial use only
- Self-hosted outputs inherit the non-commercial restriction.
VoiceStudio is AGPL-3.0 and doesn't relicense model weights. Licence guide
Related
From the VoiceStudio app at 0834c8b. 27 models listed.