ComfyUI Node
VOID Inference
Run VOID Pass 1 inpainting + optional Pass 2 refinement. Both outputs are always connected — Pass 2 mirrors Pass 1 when disabled.
VOID Inference
- void_pipeline
- video
- quadmask
- Pass 1
- Pass 2
◄prompt►
◄negative_promptThe video is not of a high quality, it has a low resolution. Watermark present in each frame. The background is solid. Strange body and strange trajectory. Distortion. ►
◄width672►
◄height384►
◄num_frames85►
◄num_inference_steps50►
◄seed42►
◄guidance_scale1.0►
◄denoise_strength1.00►
◄enable_pass2true►
◄pass2_guidance_scale6.0►
◄warped_noise_path►
CategoryVOID
Inputs (15)
| Name | Type | Default | Description |
|---|---|---|---|
| void_pipeline | VOID_PIPELINE | — | |
| video | IMAGE | Input video [T, H, W, C]. Scaled down to fit within 480×720 if needed, then snapped to the nearest multiple of 16. Frame count is trimmed to (n−1) % 4 == 0, max 85. | |
| quadmask | IMAGE | Quadmask [T, H, W, C]. 0=remove 63=overlap 127=affected 255=background. | |
| prompt | STRING | Describe the scene after object removal (e.g. 'A table with a cup.'). | |
| negative_prompt | STRING | The video is not of a high quality, it has a low resolution. Watermark present in each frame. The background is solid. Strange body and strange trajectory. Distortion. | Used by both passes. Pass 2 ignores this (always runs with empty negative). |
| width | INT | 672 | Output width in pixels. Snapped to nearest multiple of 16. |
| height | INT | 384 | Output height in pixels. Snapped to nearest multiple of 16. |
| num_frames | INT | 85 | Number of frames. Trimmed to nearest (n−1) % 4 == 0, max 85. |
| num_inference_steps | INT | 5010–200 | Denoising steps. Applies to both passes. |
| seed | INT | 420–2147483647 | RNG seed. Applies to both passes. |
| guidance_scale | FLOAT | 1.01–20 | Pass 1 CFG scale. 1.0 = distillation mode (paper default). |
| denoise_strength | FLOAT | 1.000.1–1 | Pass 1 denoise strength. 1.0 = full inpaint. |
| enable_pass2 | BOOLEAN | true | Run Pass 2 refinement. When off, Pass 2 output mirrors Pass 1. |
| pass2_guidance_scale | FLOAT | 6.01–20 | Pass 2 CFG scale. 6.0 recommended. |
| warped_noise_path | STRING | Optional path to noises.npy from make_warped_noise.py (void-model repo). Seeds Pass 2 with optical-flow-warped noise for better temporal consistency. Leave empty to use random init. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| Pass 1 | IMAGE | — |
| Pass 2 | IMAGE | — |