Turn a reference into sketch art, a 3D figurine, a sticker, a holiday card, or a product poster by describing the intended subject, style, lighting, and background. Start with the most important change, review the result, then refine one detail at a time.
Edit an image
This image-to-image AI generator takes the subject, product, composition, and perspective from your reference, then rewrites season, material, lighting, background, or shooting style per your prompt. For structure preservation, GPT Image 2 — OpenAI's latest, the same family as ChatGPT image generation and DALL·E — performs the most reliably. Pin critical details with 'keep unchanged' phrasing in your prompt to reduce drift.
Edit an image
When the model supports multi-reference input, upload subject, outfit, product, background, or style as separate references. Nano Banana 2 — a Gemini 3 Flash-powered AI image generator from image — figures out each role on its own, then merges them per your prompt. Composition, outfit swap, style harmonization, and scene transfer all happen in one render.
Edit an image
Consistent portraits and brand visuals. This page highlights popular models such as GPT Image 2, Nano Banana 2, Seedream 5.0, and Seedream 3.5 Pro. Different models are better at multi-image fusion, photoreal repainting, product edits, or portrait consistency.
Edit an imageModel availability changes. The editor is the source of truth for the options and credit estimate available before each generation.
| Capability | What to expect | Basis |
|---|---|---|
| Starting input | One or more source images, subject to the selected model | Live editor upload controls |
| Transformation direction | A prompt stating what to change and what to preserve | Provider guidance |
| Reference count, resolution, quality, and cost | Use the choices and estimate currently shown in the editor | Live editor controls |
Everything you need to know before you start creating.
It uses an uploaded image as visual reference, then applies the changes you describe in the prompt.
You can ask the model to retain the subject, pose, composition, colors, or overall style. Clear preservation instructions make the result more consistent.
It is especially useful for restyling, product-photo improvements, background changes, relighting, poster variations, and character or brand consistency.
Use a clear, unobstructed reference and explicitly name the face, outfit, product shape, color, or traits that should remain unchanged.
The editor accepts JPEG, PNG, and WebP images. It validates file size and dimensions before generation and shows an error if an upload does not meet the selected model's requirements.
Some models accept multiple reference images and others accept fewer. The editor updates the available upload slots for the selected model. Give each reference a clear role and avoid combining images with conflicting identities or styles.
Image to Image uses uploaded visual material as the starting context for a transformation. Text to Image creates a new image directly from a written description when you do not need to preserve a source subject, composition, or style.
No. References and preservation instructions improve continuity, but faces, hands, logos, readable text, patterns, and small product details can still change. Review important details before publishing the result.
Imagine Scenes shows the estimate before generation. Pricing depends on the selected model and its available output settings, such as resolution or quality. Check the updated estimate after switching models or settings.
Yes. Download the image, continue refining it here, or send it to Image to Video when you are ready to add motion.
Select a secure payment provider to continue.
Create amazing AI-generated videos and images
Create a new account with Google or GitHub to claim them. Unused credits expire 30 days after signup.
By continuing, you agree to the terms and privacy policy.