ComfyUI Node
🎛️ ThinkSound Sampler
A ComfyUI node in ThinkSound with 11 inputs and 1 output.
🎛️ ThinkSound Sampler
- thinksound_model
- feature_utils
- video
- audio
◄duration8.0►
◄steps24►
◄cfg_scale5.0►
◄seed0►
◄caption►
◄cot_description►
◄force_offloadtrue►
◄performance_modebalanced►
CategoryThinkSound
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| thinksound_model | THINKSOUND_MODEL | — | |
| feature_utils | THINKSOUND_FEATUREUTILS | — | |
| duration | FLOAT | 8.01–30 | Duration of generated audio in seconds |
| steps | INT | 241–100 | Number of denoising steps (more = better quality, slower) |
| cfg_scale | FLOAT | 5.01–20 | Classifier-free guidance scale (higher = more faithful to text) |
| seed | INT | 00–18446744073709550000 | Random seed for reproducible results |
| caption | STRING | Short description of desired audio (e.g., 'dog barking', 'ocean waves') | |
| cot_description | STRING | Detailed chain-of-thought description for enhanced audio generation | |
| force_offload | BOOLEAN | true | Offload models after generation to save VRAM |
| performance_mode | COMBO | balanced | Generation performance profile |
| videoopt | IMAGE | Input video frames for video-to-audio generation (optional) |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| audio | AUDIO | — |