ComfyUI Node
Krea2 Regional Builder (canvas)
A ComfyUI node in conditioning/krea2_regional with 18 inputs and 8 outputs.
Krea2 Regional Builder (canvas)
- clip
- image
- prev_regions
- extra_base_loras
- regions
- base_conditioning
- base_loras
- masks
- base_prompt
- region_prompts
- width
- height
◄width1024►
◄height1024►
◄grow_px0►
◄feather_px0►
◄base_prompt►
◄background►
◄aesthetics►
◄lighting►
◄medium►
◄region_append►
◄import_modewhen empty►
◄regions_data►
◄layout_in_baseposition hints►
◄import_json—►
Categoryconditioning/krea2_regional
Inputs (18)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | — | |
| width | INT | 102464–16384 | — |
| height | INT | 102464–16384 | — |
| grow_px | INT | 00–512 | — |
| feather_px | INT | 00–512 | — |
| base_prompt | STRING | Global scene description. Filled from an imported caption. | |
| background | STRING | Background description. Filled from an imported caption. | |
| aesthetics | STRING | — | |
| lighting | STRING | — | |
| medium | STRING | — | |
| region_append | STRING | Appended to every region prompt. | |
| import_mode | COMBO | when empty | 2 options: when empty, always |
| regions_data | STRING | — | |
| layout_in_base | COMBO | position hints | Put the region layout INTO the base prompt so the model composes subjects where your boxes are. Krea 2's Qwen3-VL encoder reads layout guidance; the attention masks then handle isolation/LoRAs. 'position hints' appends natural-language placements; 'full JSON' uses an Ideogram-style structured prompt as the base. |
| import_jsonopt | STRING | — | |
| imageopt | IMAGE | — | |
| prev_regionsopt | KREA2_REGIONS | — | |
| extra_base_lorasopt | KREA2_LORAS | — |
Outputs (8)
| Name | Type | Description |
|---|---|---|
| regions | KREA2_REGIONS | — |
| base_conditioning | CONDITIONING | — |
| base_loras | KREA2_LORAS | — |
| masks | MASK | — |
| base_prompt | STRING | — |
| region_prompts | STRING | — |
| width | INT | — |
| height | INT | — |