Instructions to use Zyphra/ZONOS3-Mini with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Zonos
How to use Zyphra/ZONOS3-Mini with Zonos:
# pip install git+https://github.com/Zyphra/Zonos.git import torchaudio from zonos.model import Zonos from zonos.conditioning import make_cond_dict model = Zonos.from_pretrained("Zyphra/ZONOS3-Mini", device="cuda") wav, sr = torchaudio.load("speaker.wav") # 5-10s reference clip speaker = model.make_speaker_embedding(wav, sr) cond = make_cond_dict(text="Hello, world!", speaker=speaker, language="en-us") codes = model.generate(model.prepare_conditioning(cond)) audio = model.autoencoder.decode(codes)[0].cpu() torchaudio.save("sample.wav", audio, model.autoencoder.sampling_rate) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- llama.cpp
How to use Zyphra/ZONOS3-Mini with llama.cpp:
Install (macOS, Linux)
curl -LsSf https://llama.app/install.sh | sh # Start a local OpenAI-compatible server with a web UI: llama serve -hf Zyphra/ZONOS3-Mini # Run inference directly in the terminal: llama cli -hf Zyphra/ZONOS3-Mini
Install from WinGet (Windows)
winget install llama.cpp # Start a local OpenAI-compatible server with a web UI: llama serve -hf Zyphra/ZONOS3-Mini # Run inference directly in the terminal: llama cli -hf Zyphra/ZONOS3-Mini
Build from source code
git clone https://github.com/ggml-org/llama.cpp.git cd llama.cpp cmake -B build cmake --build build -j --target llama-app # Start a local OpenAI-compatible server with a web UI: ./build/bin/llama serve -hf Zyphra/ZONOS3-Mini # Run inference directly in the terminal: ./build/bin/llama cli -hf Zyphra/ZONOS3-Mini
Use Docker
docker model run hf.co/Zyphra/ZONOS3-Mini
- Llama new
- LM Studio
- Jan
- Ollama
How to use Zyphra/ZONOS3-Mini with Ollama:
ollama run hf.co/Zyphra/ZONOS3-Mini
- Unsloth Desktop
- Docker Model Runner
How to use Zyphra/ZONOS3-Mini with Docker Model Runner:
docker model run hf.co/Zyphra/ZONOS3-Mini
- Lemonade
How to use Zyphra/ZONOS3-Mini with Lemonade:
Pull the model
# Download Lemonade from https://lemonade-server.ai/ lemonade pull Zyphra/ZONOS3-Mini
Run and chat with the model
lemonade run user.ZONOS3-Mini-{{QUANT_TAG}}List all available models
lemonade list
- Atomic Chat
Urdu is listed as supported, but Urdu TTS generation is not working correctly
Urdu is among the 43 supported languages however while testing the model for Urdu I have not been able to obtain satisfactory Urdu speech.
Hi ! Thank you for your interest in the model.
The model was trained on Urdu speech, but we did not evaluate the final model's performance on this language in ZTTS2-Eval.
I ran a small internal evaluation over Urdu speech, and found the model does somewhat well, achieving 15.3% WER.
Could you let me know how exactly the model is failing for you ?
Thanks !
George
its cloning the voice quite well but I'm noticing some inconsistencies in Urdu word pronunciation certain words don't sound natural or are not pronounced correctly which is affecting the overall quality.
We're working on a finetuning framework which should let you adapt the model in such a way to improve performance for a particular language; I'll post here once it's available.
im waiting for it, thank you