ComfyUI Node
Krea2 Edit (grounded encode)
Encodes the edit instruction grounded on the source image (training-matched semantic path).
Krea2 Edit (grounded encode)
- clip
- image
- image_b
- CONDITIONING
◄prompt►
◄grounding_px768►
◄system_prompt►
Categorykrea2edit
Inputs (6)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | — | |
| prompt | STRING | — | |
| imageopt | IMAGE | — | |
| image_bopt | IMAGE | 2nd reference (subject) for multi-ref LoRAs; vision blocks in training order: scene, subject | |
| grounding_pxopt | INT | 7680–4096 | cap longest side fed to Qwen3-VL; 0 = native |
| system_promptopt | STRING | advanced (optional): override the grounding system prompt (empty = training default). Steers what the vision encoder attends to, e.g. facial identity detail. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| CONDITIONING | CONDITIONING | — |