Source-led image generation
Transform a photo with image-to-image AI
Upload an existing image, separate the details that must remain fixed from those that can change and generate a new visual direction from the same source.
Example directions you can create
Modern minimal — Light facade, dark trim and restrained landscaping. 
Dynamic shōnen — Stronger linework, dramatic light and energetic framing. 
Travel journal — Layered memories, map-like shapes and warm paper texture.
How Transform a photo with image-to-image AI works
Image-to-image AI uses an uploaded reference to guide the new output. The source can communicate identity, product shape, camera angle, silhouette, palette or composition more directly than text alone.
A reference improves control but does not guarantee perfect identity or pixel-level consistency. Decide what is fixed and what is flexible before generating, then compare the result with the same source rather than judging style alone.
What you can create
- Multiple directions from one source — Use the same reference to explore cinematic, illustrated, editorial, professional or other supported transformations.
- More visual guidance than text alone — Communicate subject, silhouette, angle and composition through the uploaded image.
- Explicit control of what may change — Separate protected identity or geometry from flexible background, lighting and styling.
Popular use cases
- Portrait restyling — Transform a portrait into another photographic or illustrated direction while checking facial identity.
- New scene from an existing subject — Keep the main subject and explore another environment, season or lighting setup.
- Professional presentation — Use a casual source as guidance for a more controlled background, crop and visual tone.
- Product scene exploration — Keep product shape and color as reference while testing contextual photography concepts.
- Composition-led variation — Use a strong layout or camera angle as the basis for a new visual execution.
What makes a strong image-to-image reference?
The most useful reference clearly shows the subject, camera angle, proportions or composition you want the model to understand. It does not need elaborate styling if the task is to change that styling.
Avoid tiny, heavily compressed or cluttered files. Use a sharp image with even light and little obstruction when a face or product must remain recognizable.
- Clear subject and silhouette
- Useful camera angle and crop
- Enough resolution for important details
- Limited obstruction around the focal point
Identity preservation is guidance, not a guarantee
A portrait reference can guide facial structure, age cues and hairstyle, but stylization, new poses and dramatic changes may reduce resemblance. Product references can also drift in labels, logos, proportions and fine geometry.
When identity or fidelity matters, keep the transformation narrower, protect the important details in the prompt and compare every result with the uploaded source.
Image-to-image quality check: fixed versus flexible details
Return to the brief and compare every detail marked as fixed with the reference. Check identity, silhouette, product geometry, object count, camera angle and composition before judging the new style or background.
Inspect areas invented outside the original crop as well as faces, hands, reflections and text. When too many fixed elements drift, simplify the transformation instead of adding more instructions to the same prompt.
- Compare identity and silhouette
- Verify product shape and object count
- Check newly invented image areas
- Narrow the next change when structure drifts
Use text-to-image for a completely new scene
Choose the AI Image Generator when no particular person, product or composition needs to be preserved. Text-to-image allows more invention because it starts from a written brief rather than a visual reference.
Use image-to-image when the source contains important visual information that would be difficult or inefficient to describe with words alone.
How to get started
- Upload a useful reference image — Choose a clear source that shows the subject, proportions, angle or composition you need.
- Separate fixed and flexible details — State what must remain recognizable, then describe the scene, style or light that may change.
- Compare every output with the source — Check identity, silhouette, product geometry, object count and newly invented areas before refining.
Frequently asked questions
- What does image-to-image AI use from my photo?
- The reference can guide identity, shape, angle, silhouette, palette or composition while your prompt describes what should change.
- Will the result preserve identity perfectly?
- No. A reference improves guidance but cannot guarantee exact identity. Compare facial features, age cues, hair and proportions with the source.
- When should I use text-to-image instead?
- Use text-to-image when you want a completely new scene and do not need to preserve a particular person, product or composition.
- Do I need an account and what output options are available?
- An account is required before generation. Free image results include a watermark. Paid image options are watermark-free and include 2K or 4K where shown in Studio; earlier watermarked files remain unchanged.
Upload an image to transform with AI
Choose one clear reference, define fixed and flexible details and compare the generated result with the source.