ComfyUI Node
OpenAI Image Edit
A ComfyUI node in ERPK/OpenAI with 12 inputs and 1 output.
OpenAI Image Edit
- image
- client
- mask
- image
◄prompt►
◄seed-1►
◄modelgpt-image-2►
◄size1024x1024►
◄qualityauto►
◄moderationauto►
◄n1►
◄backgroundauto►
◄input_fidelityauto►
CategoryERPK/OpenAI
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| image | IMAGE | Input image(s) to edit. A single image is the default. If you pass a batched IMAGE (e.g. from an Image Batch / Rebatch node), all N images are sent as reference images — gpt-image-2 uses this for character/scene continuity (up to 16 images). Mask (if provided) applies to the first. | |
| prompt | STRING | Description of how to modify the image | |
| seed | INT | -1-1–2147483647 | Seed for reproducible outputs (best-effort). Randomizes by default. |
| clientopt | OPENAI_API_CLIENT | OpenAI API client from OpenAI API Config node | |
| maskopt | MASK | Optional mask indicating areas to edit (white = edit, black = keep) | |
| modelopt | COMBO | gpt-image-2 | Image editing model. gpt-image-2 is the latest flagship (multilingual text, character continuity across edits). gpt-image-1.5 / gpt-image-1 / gpt-image-1-mini remain available. |
| sizeopt | COMBO | 1024x1024 | Output image size. Options auto-filter based on the selected model. Edit endpoint supports the 1024-series sizes (and auto) across all GPT Image models. |
| qualityopt | COMBO | auto | Image quality (gpt-image-1 only) |
| moderationopt | COMBO | auto | Content moderation level. 'auto' uses OpenAI's default safety filters; 'low' relaxes them for permissive content. |
| nopt | INT | 11–10 | Number of edited image variants to return per call (OpenAI supports 1-10 for all GPT Image models). |
| backgroundopt | COMBO | auto | Background type for the edited output (GPT Image models only). 'transparent' requires an output format that supports transparency. |
| input_fidelityopt | COMBO | auto | Fidelity to the original input image(s). 'high' preserves details more aggressively; 'low' gives the model more creative freedom. Ignored by gpt-image-2 (always processes at high fidelity). |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |