GPT Image 2

text / image to image

Inputs:
- Text prompt
- Source image for editing
- Image references (optional)

Parameters:
- Image Size: Match Canvas, Auto, 16:9 Landscape, 9:16 Portrait, 4:3 Landscape, 3:4 Portrait, 1:1 Square HD, or 1:1 Square
- Quality: Low, Medium, or High; High is the default and affects credit cost
- Mask: optional for Edit Image; painted/white areas regenerate and clean/black areas are preserved
- Output: one PNG (fixed in Kepler)

Limits:
- Custom canvas sizes are normalized to valid width and height values before submission
- Width and height must be multiples of 16
- Maximum edge: 3840 px
- Aspect ratio limit: 3:1
- Pixel range: 655,360 to 8,294,400 total pixels
- Output count is fixed to one generated image in Kepler
- Mask dimensions must match the source image

Tips:
- Use GPT Image 2 when the result should match the project canvas size
- Use it for high-quality still generation, source-image edits, visual variations, backgrounds, and final-ready image assets
- Choose Match Canvas for timeline shots and explicit presets for standalone images
- Generated output is stored as new media and placed into the editor without overwriting the original source.