ComfyUI Node
Krea2 Two Character Composer
Two regional prompts and character LoRAs for Krea 2. Split the image left/right or top/bottom, place two free boxes or connect masks. Negative prompts use the NegPiP Prompt pack when it is installed.
Krea2 Two Character Composer
- model
- clip
- latent
- character_1_mask
- character_2_mask
- model
- positive
- latent
- region_masks
- negative
- prompts
◄base_prompt►
◄negative_prompt►
◄character_1_trigger►
◄character_1_prompt►
◄character_1_loraNone►
◄character_1_strength1.00►
◄character_2_trigger►
◄character_2_prompt►
◄character_2_loraNone►
◄character_2_strength1.00►
◄directionLeft / Right►
◄character_1_percent50►
◄separationBalanced►
◄boxes5, 10, 42, 85; 53, 10, 42, 85►
◄lora_edge_softness0►
◄base_strength1.00►
◄preview_ratioAuto►
◄width—►
◄height—►
Categoryconditioning/krea2_regional
Inputs (24)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| latent | LATENT | — | |
| base_prompt | STRING | Read by the whole image. A placement sentence for both characters is added automatically. | |
| negative_prompt | STRING | Compiled to negative weights by the installed NegPiP Prompt pack (comfyui-negpip-zimage). Works at CFG 1. Leave empty if you don't use NegPiP. | |
| character_1_trigger | STRING | Goes first in this character's prompt and is kept out of the shared scene prompt, so Krea 2 doesn't render it as text. | |
| character_1_prompt | STRING | Only this region reads this prompt. With NegPiP installed, negative weights like (beard:-0.6) apply to this character only. | |
| character_1_lora | COMBO | None | 1 options: None |
| character_1_strength | FLOAT | 1.00-10–10 | — |
| character_2_trigger | STRING | Goes first in this character's prompt and is kept out of the shared scene prompt, so Krea 2 doesn't render it as text. | |
| character_2_prompt | STRING | Only this region reads this prompt. With NegPiP installed, negative weights like (beard:-0.6) apply to this character only. | |
| character_2_lora | COMBO | None | 1 options: None |
| character_2_strength | FLOAT | 1.00-10–10 | — |
| direction | COMBO | Left / Right | Split left/right or top/bottom, or place two Free boxes. A connected character mask replaces that character's area. |
| character_1_percent | INT | 501–99 | Image area for Character 1 (left or top). Character 2 gets 100 minus this value. |
| separation | COMBO | Balanced | Balanced preserves scene interaction. Strong also restricts cross-region image attention for the first half of sampling. Soft (experimental) lets the whole image read both character prompts for the last 40% of sampling. |
| boxesopt | STRING | 5, 10, 42, 85; 53, 10, 42, 85 | Free boxes: 'x, y, width, height; x, y, width, height' in percent of the image, Character 1 first. Drag them in the preview. Character 1 is in front where boxes overlap. |
| lora_edge_softnessopt | INT | 00–256 | Pixels over which the two character LoRAs fade into each other at their boundary. The prompts stay separated. 0 = hard edge. |
| base_strengthopt | FLOAT | 1.000.05–2 | Experimental. How strongly the characters' areas follow the base prompt compared to their own prompt. 1.0 = unchanged. |
| character_1_maskopt | MASK | Optional. Replaces Character 1's area. Of a batch, the first mask is used. | |
| character_2_maskopt | MASK | Optional. Replaces Character 2's area. Of a batch, the first mask is used. | |
| preview_ratioopt | COMBO | Auto | Preview only: choose a fixed ratio before sampling, or Auto to read connected width/height or a recognizable latent resolution. Generation dimensions come from the LATENT. |
| widthopt | INT | Optional preview width. Connect together with height. Does not resize the latent. | |
| heightopt | INT | Optional preview height. Connect together with width. Does not resize the latent. |
Outputs (6)
| Name | Type | Description |
|---|---|---|
| model | MODEL | Patched Krea 2 model. Connect directly to the sampler. |
| positive | CONDITIONING | Regional positive conditioning. |
| latent | LATENT | The input latent, passed through. |
| region_masks | MASK | The two region masks as a batch, for previewing. |
| negative | CONDITIONING | Matching empty negative for the sampler (only used at CFG > 1). |
| prompts | STRING | The scene and character prompts exactly as encoded. |