ComfyUI Node
⚗ Alchemica Face Blend (2-6 photos, per-trait guide)
Blends 2-6 people into ONE new person: every face trait is an in-between of the photos (skin, eye colour, jaw...), by overall weights and a per-trait guide. Output = a description for the prompt.
⚗ Alchemica Face Blend (2-6 photos, per-trait guide)
- clip
- image_1
- image_2
- image_3
- image_4
- image_5
- image_6
- subject
- report
◄weight_11.00►
◄weight_21.00►
◄weight_31.00►
◄weight_41.00►
◄weight_51.00►
◄weight_61.00►
◄trait_guide►
◄idea►
◄words120►
◄body_from_photostrue►
◄reread_photosfalse►
◄skin_compensationtrue►
Category⚗ AlchemicaMente/characters
Inputs (19)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | Krea 2's text encoder, best straight from the loader (without LoRAs) | |
| image_1 | IMAGE | — | |
| image_2 | IMAGE | — | |
| weight_1 | FLOAT | 1.000–4 | How much photo 1 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| weight_2 | FLOAT | 1.000–4 | How much photo 2 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| weight_3 | FLOAT | 1.000–4 | How much photo 3 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| weight_4 | FLOAT | 1.000–4 | How much photo 4 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| weight_5 | FLOAT | 1.000–4 | How much photo 5 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| weight_6 | FLOAT | 1.000–4 | How much photo 6 pulls overall. Normalised (2 and 1 = two thirds and one third). 0 = ignore the photo |
| trait_guide | STRING | One line per trait, overrides the weights for that trait: occhi = 2 (eyes from photo 2) mascella = 1:70 2:30 pelle, capelli = 1 Traits: origine, viso, occhi, forma occhi, colore occhi, naso, labbra, mascella, zigomi, sopracciglia, capelli, colore capelli, pettinatura, pelle, segni, fronte, eta, corporatura, corpo | |
| idea | STRING | Pose, body, clothing, mood. It never changes the face and hair from the photos | |
| words | INT | 12050–220 | Length of the description (17 traits: 110-140) |
| body_from_photos | BOOLEAN | true | Also blend build and body shape from the photos. Off = the body comes only from the idea |
| reread_photos | BOOLEAN | false | Photos already read are remembered on disk (fast). On = read them again (e.g. after changing the text encoder) |
| skin_compensation | BOOLEAN | true | Krea 2 paints skin lighter than the words say. On = picks the word by a table MEASURED on a fine-tuned NVFP4 checkpoint (ITA of the painted skin), so the image matches the blend of the photos. Off = plain words (other checkpoints) |
| image_3opt | IMAGE | — | |
| image_4opt | IMAGE | — | |
| image_5opt | IMAGE | — | |
| image_6opt | IMAGE | — |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| subject | STRING | — |
| report | STRING | — |