GPT Image 2 Edit
Take an image you already have and move it forward — swap a background, refine a product, adapt a reference — by describing the change in plain language, up to 4K.
Upload what you have and say what to adjust — no masks to paint, no blank canvas to fill.
Tell the model what to swap and what to leave alone, in ordinary language.
Pick 1K, 2K, or 4K and a quality setting for the adapted result.
- Text and image-to-video prompts
- 480p, 720p, and 1080p output options
- Prompt ideas from uploaded showcase media
- Standard Seedance video mode
- 480p, 720p, and 1080p output options
- Best for standard Seedance 2 generation
- 480p and 720p generation modes
- Lower-credit prompt testing
- Same text and image entry points
- 480p and 720p generation modes
- Lowest credit cost in the lineup
- Text and image input
- Text-to-image generation
- 1K, 2K, and 4K image options
- Useful for style frames and references
- Image editing with references
- 1K, 2K, and 4K output options
- Useful before image-to-video drafts
GPT Image 2 Edit
What is GPT Image 2 Edit?
GPT Image 2 Edit takes an image you already have and changes it on your terms. Upload one or more source images, describe what to swap and what to keep, and it returns an adapted version at up to 4K. It always works from an existing picture — this is for iterating on something real, not conjuring a new image from a blank prompt.
Upload a source image and name the change: what should shift, what should stay, and how you will use the result.
Bring in several reference images at once when the edit needs product, style, or subject detail from more than one source.
Set quality (low, medium, high) and resolution (1K, 2K, 4K) for the adapted output.
Creative engine
Editing from a source image
Upload, describe the change, choose settings, and read the credit estimate before you generate.
Changes you can make in place
Targeted adjustments that move one part of the picture forward while the rest stays exactly as you left it.
New background, same subject
Hold the subject and rebuild the backdrop, light, and shadow to suit a different setting.
Why edit instead of regenerate?
Choose Edit when the image already mostly works and you need controlled, deliberate changes — not a fresh roll of the dice.
Hold onto what works
Keep a product's shape, a face, a pose, or the composition while you change only the parts that should move.
Describe, don't paint
Spell out what changes and what stays in plain language rather than masking selections by hand.
Get a reference ready
Tidy a still — background, framing, subject — before it becomes Seedance image-to-video input.
Plain-language editing, up to 4K
The core of how GPT Image 2 Edit reshapes an image you already have.
Image-to-image editing
Begin from uploaded reference images and steer the change with written instructions.
Several references at once
Pull in multiple source images when the edit needs product, style, or character context.
Quality you choose
Pick low, medium, or high to weigh detail against credit cost.
Up to 4K
Return the adapted image at 1K, 2K, or 4K depending on where it goes next.
GPT Image 2 Edit technical specs
The parameters that define the editing model.
- Model
- GPT Image 2 Edit (gpt-image-2-edit)
- Required input
- One or more reference images
- Edit instruction
- Natural-language prompt (up to 20,000 characters)
- Quality
- Low, medium, high
- Resolution
- 1K, 2K, 4K
- Output
- One edited image per request
Upload, describe, adapt
Four steps from a source image to an adapted one.
Upload what you have
Start with the source image, or images, that should guide the change.
Name the change
Say what to swap, what to keep, and where the adapted image will be used.
Choose settings
Set quality, resolution, and output format before you run the edit.
Check cost and generate
Read the credit estimate, run the edit, and download the adapted image.
What to adapt with GPT Image 2 Edit
Reach for Edit whenever you already have an image that needs a targeted, deliberate change.
Tidying product shots
Swap the background, light, or scene while the product stays clearly itself.
Thumbnail variations
Build alternate takes on a poster, frame, or social image from one source.
Prepping video references
Refine a still before it goes into Seedance as image-to-video input.
Shifting style or mood
Retune the look, color, or setting without going back to a blank prompt.
GPT Image 2 Edit FAQ
Questions people ask about adapting images with AI.
Adapt your image with GPT Image 2 Edit
Upload your image, describe the change, choose settings, and read the credit estimate before you generate.
Explore video and image models
Seedance 2.5
Cinematic Seedance video path with text and image prompts, prompt ideas, and credit preview.
Seedance 2
Standard Seedance video mode for 480p, 720p, and 1080p generation.
Seedance 2 Fast
A faster, lower-cost Seedance video option for 480p and 720p drafts.
Seedance 2.0 Mini
Low-cost lightweight Seedance tier for 480p and 720p video at the lowest credit cost.
GPT Image 2
Create reference images, style frames, thumbnails, and product visuals for image or video projects.