ComfyUI Node Runs on cloud
Revise Prompt
Rewrite a text-to-image prompt from a judge's notes so the next attempt shows what the last one missed, keeping the prompt's subject and style. Without notes the prompt passes through as written, so a retry loop's first pass uses it unchanged.
Revise Prompt
- clip
- last_image
- prompt
◄prompt►
◄temperature0.00►
◄seed0►
◄max_length384►
◄feedback—►
◄last_prompt—►
CategoryWAS Suite/Text
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| clip | CLIP | qwen3.5_9b_int8_convrot.safetensors from Load CLIP with type krea2, or any language model CLIP. | |
| prompt | STRING | The prompt as written, as 'three red apples on a blue plate'. | |
| temperature | FLOAT | 0.000–2 | 0 = the likeliest words, the same revision every time; 0.7 = varied wording per seed. |
| seed | INT | 00–18446744073709550000 | The seed sampling draws from. Ignored at temperature 0. |
| max_length | INT | 38432–2048 | The most tokens the revised prompt may run to: 384 covers about 280 words. |
| feedbackopt | STRING | Judge Images' notes on the last attempt, as 'only two apples, the plate is white'. Unwired or empty = the prompt passes through as written. | |
| last_promptopt | STRING | The prompt the last attempt was made from, as 'three red apples on a blue plate'. | |
| last_imageopt | IMAGE | The last attempt's picture, as Judge Images' best. Read only by a model that sees pictures. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| prompt | STRING | The prompt for the next attempt, for CLIP Text Encode. |