GPT-IMG
Preserve the subject, rewrite the scene
Keep a product, person, object, or layout recognizable while changing the lighting, setting, surface material, background, or mood.
GPT-IMGGPT image editor
Start from a picture you already have. Upload it, describe the change in a sentence, and get a version back with the parts you did not mention left alone.
Upload a photo and say what should change. The GPT image editor runs GPT Image 2.5, Nano Banana, and more, with the credit cost shown per edit.
Reference GenerationStyle RedrawSubject RetentionMulti-Image Reference
148 / 20,000
Image
Upload at least one reference image before generating.
GPT image editor at a glance
Updated
- Free to start
- A new account gets 30 credits and needs no payment details. One 1K Standard edit on GPT Image 2.5 costs 15 credits, so the first two edits are free. Images made on free credits carry a watermark.
- Pay as you go
- Credit packs are one-time and never expire: $4.99 buys 300 credits (20 edits at that setting) and $26.99 buys 1,800 (120 edits). Buying credits or subscribing removes the watermark.
- Models that take an upload
- GPT Image 2.5, GPT Image 2, Nano Banana 2.1, Nano Banana Pro, Seedream 5.0 Pro, FLUX 2 Pro, and Qwen Image Edit. GPT Image 2.5 accepts up to five reference photos in one run.
- How you describe an edit
- Write one sentence about what should change. Parts of the photo you did not mention are meant to stay as they are, though the whole frame is re-rendered rather than patched.
- If something goes wrong
- A failed job is refunded automatically. Every version is saved as its own file in History, so a bad round never overwrites a good one.
Models that accept a reference image
Not every model here reads an upload. These do, and each one interprets "change this, keep that" a little differently.
GPT Image 2.5
OpenAI latest image model. Sharper instruction following, cleaner in-image text, up to 4K.
Nano Banana 2.1
Google latest image model. Sharper edits, cleaner in-image text, steadier subject consistency, up to 4K.
Nano Banana Pro
Highest quality Google model. Supports up to 4K resolution.
GPT Image 2
Most capable OpenAI model. Supports all quality tiers and up to 4K.
Seedream 5.0 Pro
Strong typography and poster layouts. Handles dense in-image text at 1K and 2K.
FLUX 2 Pro
Photorealistic detail with tight prompt control. Accepts up to 8 reference images.
Qwen Image Edit
Edit an existing image by describing the change. Requires a reference image.
From one reference image to controlled variants
Image-to-image generation is useful when a text prompt alone is not enough. The uploaded image supplies structure, subject context, style direction, or edit constraints.
GPT-IMG
Turn templates into repeatable edit flows
Start from gallery prompt patterns for common work such as ecommerce photos, professional headshots, thumbnails, and room redesign concepts.
GPT-IMG
Combine prompt detail with visual reference
Use the prompt to define what should change and the reference image to define what should remain consistent.
GPT-IMG
Track variants through the same generation system
Reference-image jobs use the same login, credits, pending state, history, and refund chain as prompt-only generation.
GPT-IMG
Swapping the context, keeping the subject
Move a product onto a different surface, change the season outside the window, or relight a portrait, without redrawing the thing itself.
GPT-IMG
Restyling without losing the likeness
Change art direction, material, or palette while the subject stays recognizable across the whole set of variants.
GPT-IMG
Several rounds on the same image
Feed an output back in as the next input. GPT Image 2.5 in particular holds the subject steady across repeated passes.
GPT-IMG
Edits you can undo by ignoring them
Every version is a separate entry in History with its own file, so a bad round costs credits but never overwrites the good one.
Keep exploring GPT-IMG
Move between prompt-only generation, reference edits, gallery templates, and model-specific pages without changing the underlying workflow.
Text to Image
Start from words only when you do not need an uploaded image to guide subject, structure, or style.
Edit Text in Image
The narrower version of this page: name the words to swap and leave the rest of the shot alone.
GPT Image 2.5
The default editing model, and the one that holds a subject steady across several rounds of revision.
Multiple reference image generator
The multi-photo version of this page: give each of up to five references a job, such as person, product, or style.
Ideogram 4.5 editing test
A three-edit chain on Ideogram 4.5 and Nano Banana 2.1, with prompts, credit cost, and a drift measurement.
Choose Your Plan
Upgrade for larger monthly credits, faster generation, and commercial-ready output.
Need more without a subscription?
Buy credits once and use them whenever you need. Purchased credits never expire.
FAQs
Practical details for this generator page.
Start with GPT image editor
Use the prompt panel above with the same account, credits, and history workflow.