ComfyUI Node
OmniVoice Generate Audio
A ComfyUI node in OmniVoice/Process with 18 inputs and 2 outputs.
OmniVoice Generate Audio
- pipe
- audio
- status
◄textHello from OmniVoice.►
◄languageauto►
◄style_gendernone►
◄style_agenone►
◄style_pitchnone►
◄style_accentnone►
◄num_step32►
◄guidance_scale2.0►
◄t_shift0.10►
◄layer_penalty_factor5.0►
◄position_temperature5.0►
◄class_temperature0.00►
◄speed1.00►
◄seed0►
◄use_durationfalse►
◄duration10.0►
◄postprocess_outputtrue►
CategoryOmniVoice/Process
Inputs (18)
| Name | Type | Default | Description |
|---|---|---|---|
| pipe | OMNIVOICE_PIPE | Pipe output from OmniVoice Load Model | |
| text | STRING | Hello from OmniVoice. | Target speech text |
| language | COMBO | auto | Language hint for TTS; auto lets model infer language |
| style_gender | COMBO | none | no ref audio only |
| style_age | COMBO | none | no ref audio only |
| style_pitch | COMBO | none | no ref audio only |
| style_accent | COMBO | none | no ref audio only |
| num_step | INT | 324–128 | Diffusion sampling steps (higher = slower, often cleaner) |
| guidance_scale | FLOAT | 2.00–20 | Classifier-free guidance strength |
| t_shift | FLOAT | 0.100–1 | Diffusion timestep shift; lower tends to favor low-SNR detail |
| layer_penalty_factor | FLOAT | 5.00–20 | Penalty encouraging earlier codebook layers to unmask first |
| position_temperature | FLOAT | 5.00–20 | Temperature for position selection during generation |
| class_temperature | FLOAT | 0.000–2 | Token class sampling temperature (0 = greedy) |
| speed | FLOAT | 1.000.25–4 | — |
| seed | INT | 00–2147483647 | Random seed for reproducibility |
| use_duration | BOOLEAN | false | Enable fixed output duration override |
| duration | FLOAT | 10.00–120 | Target output duration in seconds (used only when enabled) |
| postprocess_output | BOOLEAN | true | Apply output postprocessing (trim/fade/pad cleanup) |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| audio | AUDIO | — |
| status | STRING | — |