VibeVoice MLX
Collection
VibeVoice-{1.5b, 7b} converted to MLX • 2 items • Updated
VibeVoice 7B — MLX
MLX-converted fp16 weights for vibevoice/VibeVoice-7B.
For inference code, benchmarks, and documentation see vibevoice-mlx.
Quick start
git clone https://github.com/gafiatulin/vibevoice-mlx && cd vibevoice-mlx
uv sync
# Basic synthesis (weights download automatically)
uv run vibevoice-mlx --model gafiatulin/vibevoice-7b-mlx --text "Hello, world!" --output hello.wav
# Voice cloning with INT8 quantization
uv run vibevoice-mlx --model gafiatulin/vibevoice-7b-mlx --quantize 8 \
--ref-audio speaker.wav --text "Clone this voice" --output cloned.wav
Performance
Benchmarked on M4 Max 64GB with voice cloning (~30s audio):
| Config | RTF | Gen | Peak Mem |
|---|---|---|---|
| fp16 | 0.53x | 53.0s | 21.7 GB |
| int8 | 1.06x | 29.6s | 14.9 GB |
| int4 | 1.16x | 25.8s | 11.2 GB |
| int8, no-semantic | 1.24x | 23.3s | 13.6 GB |
| int4, no-semantic | 1.37x | 19.5s | 9.8 GB |
- Downloads last month
- 255
Model size
9B params
Tensor type
BF16
·
Hardware compatibility
Log In to add your hardware
Quantized
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for gafiatulin/vibevoice-7b-mlx
Base model
vibevoice/VibeVoice-7B