GPT Image 2 Edit

Edit one or more reference images with GPT Image 2 on fal.ai, with explicit quality and output-size controls plus optional masking.

Model Overview

GPT Image 2 Edit applies prompt-guided edits to supplied images. This Nodespell surface keeps size and quality choices aligned with fal's published pricing table while still using post-run billing-event correction for the exact billed cost.

Best At

• Prompt-guided image edits from one or more reference images.
• Structured iteration where you want to pick a published size or explicit custom width and height instead of relying on vendor defaults.
• Optional mask-driven edits when only part of the input image should change.

Limitations / Not Good At

• This node fixes generation to one output image per run.
• Custom sizes still need to satisfy fal's GPT Image 2 limits: multiples of 16, max edge 3840px, aspect ratio up to 3:1, and total pixels between 655,360 and 8,294,400.
• The final billed cost can still differ from the starting estimate, so Nodespell corrects credits after completion.

Ideal Use Cases

• Product, character, and scene edits from supplied image references.
• Inpainting-style workflows that need an explicit mask input.
• Image editing flows where predictable initial estimates are worth preserving.

Input & Output Format

• Input: required image_input and prompt, optional mask_input, plus quality, image_size_preset, optional custom_width / custom_height, and output_format.
• Output: edited image asset returned on response.

Performance Notes

• Larger output sizes and higher quality settings increase cost.
• Custom sizes use the nearest published GPT Image 2 pricing bucket for the starting estimate, then Nodespell replaces that estimate with fal's billed amount after the task finishes.