ComfyUI Node
Fantastic H3 Edit Composite
Put the edit back into the original footage: the area H3 regenerated (the mask, grown and rounded out to the latent's 16-pixel cells, feather included) is pasted into the Media Loader's clip through a soft edge, and everything else is the source file's own pixels, untouched by the VAE. Wire it between VAE Decode and Create Video, with the same references the Text Encode gets. With no clip being edited it passes the frames through.
Fantastic H3 Edit Composite
- images
- references
- images
◄blend8►
◄max_size0►
Categoryconditioning/video_models
Inputs (4)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | The decoded frames from VAE Decode. | |
| references | H3_REFS | The Media Loader bundle, from the loader or the Prompt Builder's references output — the same one the Text Encode gets. | |
| blend | INT | 80–64 | Extra soft edge, in output pixels, on top of the one the latent cells give. |
| max_size | INT | 00–8192 | Cap the output's long edge. 0 keeps the source clip's size (after the loader's crop and size cap). |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| images | IMAGE | — |