ComfyUI Node
Forbidden Vision 🎯 Fixer
A ComfyUI node in Forbidden Vision with 35 inputs and 4 outputs.
Forbidden Vision 🎯 Fixer
- model
- vae
- positive
- negative
- image
- latent
- clip
- final_image
- processed_face
- side_by_side_comparison
- final_mask
◄steps10►
◄cfg_scale3.0►
◄samplereuler_ancestral►
◄schedulersgm_uniform►
◄denoise_strength0.40►
◄seed0►
◄face_selection0►
◄detection_confidence0.75►
◄manual_rotationNone►
◄processing_resolution1024►
◄enable_pre_upscaletrue►
◄upscaler_modelFast 4x (Lanczos)►
◄crop_padding1.6►
◄face_positive_prompt►
◄replace_positive_promptfalse►
◄face_negative_prompt►
◄replace_negative_promptfalse►
◄exclusions►
◄blend_softness8►
◄mask_expansion2►
◄sampling_mask_blur_size21►
◄sampling_mask_blur_strength1.0►
◄enable_color_correctiontrue►
◄enable_segmentationtrue►
◄enable_differential_diffusiontrue►
◄enable_lightness_rescuetrue►
◄enable_final_refinementtrue►
◄offload_models_to_cputrue►
CategoryForbidden Vision
Inputs (35)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| vae | VAE | — | |
| positive | CONDITIONING | — | |
| negative | CONDITIONING | — | |
| steps | INT | 101–100 | — |
| cfg_scale | FLOAT | 3.00–30 | — |
| sampler | COMBO | euler_ancestral | 44 options: euler, euler_cfg_pp, euler_ancestral, euler_ancestral_cfg_pp, heun, heunpp2, +38 |
| scheduler | COMBO | sgm_uniform | 9 options: simple, sgm_uniform, karras, exponential, ddim_uniform, beta, +3 |
| denoise_strength | FLOAT | 0.400–1 | — |
| seed | INT | 00–18446744073709550000 | — |
| face_selection | INT | 00–20 | 0=All faces, 1=1st face, etc. |
| detection_confidence | FLOAT | 0.750.1–1 | Face detection confidence threshold. |
| manual_rotation | COMBO | None | Manually rotate face crop before processing |
| processing_resolution | INT | 1024512–2048 | The resolution for processing. |
| enable_pre_upscale | BOOLEAN | true | Enable to upscale small faces with an AI model before processing. |
| upscaler_model | COMBO | Fast 4x (Lanczos) | The model used for pre-upscaling small faces. |
| crop_padding | FLOAT | 1.61–3 | Padding added to the face region before inpaint. |
| face_positive_prompt | STRING | — | |
| replace_positive_prompt | BOOLEAN | false | — |
| face_negative_prompt | STRING | — | |
| replace_negative_prompt | BOOLEAN | false | — |
| exclusions | STRING | Words/tags to remove from the main prompt specifically for the face processing step. | |
| blend_softness | INT | 80–200 | — |
| mask_expansion | INT | 20–100 | — |
| sampling_mask_blur_size | INT | 211–101 | — |
| sampling_mask_blur_strength | FLOAT | 1.00.1–6 | — |
| enable_color_correction | BOOLEAN | true | — |
| enable_segmentation | BOOLEAN | true | Use AI segmentation. If disabled, creates oval masks. |
| enable_differential_diffusion | BOOLEAN | true | Better blending. At high noise, the mask allows structure changes; at low noise, it locks the background. |
| enable_lightness_rescue | BOOLEAN | true | If the generated face is darker than original, brighten it. |
| enable_final_refinement | BOOLEAN | true | Runs a quick 0.05 denoise pass at the end. Cleans artifacts and improves skin texture with sensitive models and higher denoise. Highly recommended. |
| offload_models_to_cpu | BOOLEAN | true | Move face detection/segmentation/corrector models from VRAM to RAM after processing. Frees GPU memory for other nodes (e.g. checkpoint swaps in queued workflows). Disable if you're chaining many face passes and want to skip the small reload overhead. |
| imageopt | IMAGE | Optional image input. If latent is also provided, latent will be used. | |
| latentopt | LATENT | Optional latent input. Will be decoded for processing. | |
| clipopt | CLIP | Optional: Required only if using face prompts. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| final_image | IMAGE | — |
| processed_face | IMAGE | — |
| side_by_side_comparison | IMAGE | — |
| final_mask | MASK | — |