ComfyUI Node
OmniVoice Model Loader
A ComfyUI node in OmniVoice with 3 inputs and 1 output.
OmniVoice Model Loader
- model
◄devicecuda:0►
◄dtypefloat16►
◄compilefalse►
CategoryOmniVoice
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| device | COMBO | cuda:0 | 3 options: cuda:0, cuda:1, cpu |
| dtype | COMBO | float16 | 3 options: float16, bfloat16, float32 |
| compile | BOOLEAN | false | Run torch.compile() on the model after loading. First generation will be slow (~30-60s warmup) while the graph is compiled, then every subsequent generation in the session will be faster. Recommended for audiobook pipelines. Requires PyTorch 2.0+. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| model | OMNIVOICE_MODEL | — |