Views
No views yet
| Repository | Format | Notes |
|---|---|---|
mlx-community/Zyphra-ZONOS2 | BF16 | Official MLX conversion for mlx-audio |
pip install mlx-audio1from mlx_audio.audio_io import write as audio_write
2from mlx_audio.tts import load
3
4model = load("mlx-community/Zyphra-ZONOS2", lazy=True)
5
6result = next(model.generate(
7 text="Hello, this is ZONOS two running locally with MLX audio.",
8 max_tokens=220,
9))
10
11audio_write("zonos2.wav", result.audio, result.sample_rate)ref_audio. Clean speech-only clips work best.1result = next(model.generate(
2 text="This text will be spoken with the reference speaker.",
3 ref_audio="speaker.wav",
4 max_tokens=220,
5))1speaker = model.extract_speaker_embedding("speaker.wav")
2
3result = next(model.generate(
4 text="This reuses a precomputed speaker embedding.",
5 speaker_embedding=speaker,
6 max_tokens=220,
7))1python -m mlx_audio.tts.generate \
2 --model mlx-community/Zyphra-ZONOS2 \
3 --text "Hello, this is ZONOS two running with MLX audio." \
4 --output_path outputs \
5 --file_prefix zonos21python -m mlx_audio.tts.generate \
2 --model mlx-community/Zyphra-ZONOS2 \
3 --text "This text will use the voice from the reference clip." \
4 --ref_audio speaker.wav \
5 --output_path outputs \
6 --file_prefix zonos2_clone| Parameter | Default | Description |
|---|---|---|
ref_audio | None | Reference audio path or array for voice cloning |
speaker_embedding | None | Precomputed 2048-D speaker embedding, Python API only |
max_tokens | 1024 | Maximum number of audio token frames |
temperature | 1.15 | Sampling temperature |
top_k | 106 | Top-k sampling filter |
top_p | 0.0 | Nucleus sampling filter, disabled at 0 |
min_p | 0.18 | Minimum-probability sampling filter |
repetition_penalty | 1.2 | Repetition penalty applied to recent audio tokens |
seed | None | Seed for deterministic sampling |
text_normalization | True | English text normalization toggle, Python API only |