ComfyUI Node
Nougan RefMod Apply 🎯
A ComfyUI node in Nougan/RefMod with 15 inputs and 3 outputs.
Nougan RefMod Apply 🎯
- clip
- positive
- negative
- extra_image
- vae
- positive
- negative
- latent
◄model▾►
◄refmod_1▾►
◄prompt—►
◄negative_prompt—►
◄refmod_2▾►
◄refmod_3▾►
◄extra_image_resolution1024►
◄extra_image_scale▾►
◄ref_strength1.00►
◄ref_on_negativetrue►
CategoryNougan/RefMod
Inputs (15)
| Name | Type | Default | Description |
|---|---|---|---|
| model | COMBO | auto reads the model from the RefMod file. Pick one explicitly to apply extra_image only, or to refuse files from another model. | |
| refmod_1 | COMBO | 1 options: None | |
| clipopt | CLIP | Qwen 2.1 / Fibo 2: the text encoder. | |
| promptopt | STRING | Qwen 2.1 / Fibo 2: encoded here with the references. | |
| negative_promptopt | STRING | — | |
| positiveopt | CONDITIONING | Flux 3: your encoded prompt. | |
| negativeopt | CONDITIONING | — | |
| refmod_2opt | COMBO | 1 options: None | |
| refmod_3opt | COMBO | 1 options: None | |
| extra_imageopt | IMAGE | Optional plain reference(s), prepared on the fly. | |
| vaeopt | VAE | Needed for extra_image. | |
| extra_image_resolutionopt | INT | 10240–4096 | — |
| extra_image_scaleopt | COMBO | Unused. Kept (hidden) so saved workflows keep their widget order. | |
| ref_strengthopt | FLOAT | 1.000–1 | How long the references stay active, as a fraction of the denoising steps counted from the start. 1.0 = every step. 0.6 = the references steer the first 60% of the steps, then the prompt alone finishes. 0 = references off. Lower = looser likeness and more prompt freedom. |
| ref_on_negativeopt | BOOLEAN | true | Also attach the reference latents to the negative (usual for edit models). Turn off if the result looks over-copied. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| positive | CONDITIONING | — |
| negative | CONDITIONING | — |
| latent | LATENT | — |