Upload a reference image and describe what you want to keep or change. Choose your aspect ratio and resolution, then generate in your workspace.
Explore images created by our team, including a portrait, a coffee cup, a lakeside house and a forest greenhouse. Consider which subject details, materials, colors or lighting you would like a reference to guide.
An uploaded image gives your request a visual starting point. For a living room, specify whether the layout, furniture or palette matters, then describe the materials and lighting you want to explore. If the result changes an important feature, make that priority clearer and simplify conflicting instructions. References guide generation; exact preservation is not guaranteed.
Use a product reference to communicate its appearance, and describe the surrounding scene separately: a skincare bottle on pale stone, a clean background and soft side lighting. Identify the shape, color or placement you want the result to follow. Check the generated bottle, label and lettering against the real product before using it in a campaign.
Explain how the portrait should change: softer window light, a different color palette or a new background. State which features or framing you want to retain, rather than relying on a general request to improve the photo. Review the resulting face, clothing and pose; a reference is not a guarantee of exact likeness or identity consistency.
With one reference image, explain which features should guide the result. For a courtyard concept, identify the architecture, colors or atmosphere that matter, then describe what you want to change. Use a clear, relevant picture and avoid incompatible instructions. You can also start without a reference using the AI text to image generator.
Decide how the new image will be used before choosing its aspect ratio. Describe the viewing angle, subject placement and space you need around it. For a bicycle scene, specify whether the bicycle or the surrounding landscape should dominate. The source image’s shape does not lock the output ratio. Check composition rather than treating higher resolution as a fix for misplaced subjects.
Explain the reference’s role, what matters in the original and what should change. Keep the request focused. After reviewing a result, revise one priority at a time; the instructions describe your goal, not a guarantee that every original detail will remain unchanged.
Explain how the reference should guide the result: subject appearance, palette, architecture or overall atmosphere. Choose one clear image and identify the features that matter.Use the reference image for the bottle’s appearance
Name the important characteristics you want to retain, such as shape, color, subject placement or layout. If the result drifts, make the highest-priority features explicit and remove conflicting requirements.Aim to retain the bottle’s overall shape and color
Describe the desired change in concrete terms: a new setting, material, light source or style. Keep preservation and change instructions compatible. Background or lighting requests can also affect other details in the generated image.Place it on pale natural stone with a warm background and soft side lighting
Specify the viewpoint, framing, subject position and any empty space. Select aspect ratio and resolution in the controls. If the framing is wrong, revise those instructions and the chosen ratio rather than only increasing resolution.Show the bottle near the center, with clean space around it
Use the reference image for the skincare bottle’s appearance. Aim to retain its overall shape and color. Place it on pale natural stone with a simple warm background and soft side lighting. Show the bottle near the center, with clean space around it. This teaching example illustrates the prompt structure; it is not the original generation record for the displayed image.
Add an image, explain your intended changes, then generate and check the result in your workspace.
Select one JPEG, PNG or WebP image with an original file smaller than 50 MiB. Choose a reference that clearly shows what matters.
Write your request in up to 1,000 Unicode characters and choose the aspect ratio and resolution independently. Each task costs 3 credits at 1K, 4 credits at 2K or 5 credits at Up to 4K.
Continue to your workspace and sign in to generate. Compare the completed image with your reference, then download the PNG or revise your request.
AI image to image generation uses one reference image together with a written prompt to guide a new result. Text-only generation starts from your description without an uploaded picture. See the AI image generator for an overview of both approaches.
Select one JPEG, PNG or WebP reference image with an original file smaller than 50 MiB. Before submission, your browser automatically compresses it to a static WebP smaller than 20,000 bytes (20 KB, decimal), preserving the full frame and aspect ratio without cropping. Fine text and textures may become blurred. A written prompt is still required. Describe which parts of the reference matter and the changes you want. Without a reference, the tool generates from text alone.
Not necessarily. References can guide the subject, colors, layout or style, but the result may change details you wanted to retain. State those priorities clearly and check the output, especially faces, product shapes, logos and lettering. This workflow does not provide a masking brush or a reference-strength slider.
Choose from five aspect ratios—1:1, 16:9, 9:16, 3:2 and 2:3—and 1K, 2K or Up to 4K resolution settings. All five ratios are available at every resolution tier, and aspect ratio and resolution are independent settings. The highest pixel dimensions depend on the aspect ratio. Each task produces one PNG and costs 3, 4 or 5 credits respectively. New accounts receive 50 free credits, valid for 30 days. See pricing and credit options for further use.
Yes. Free and paying users may use generated images commercially. Only upload reference material you own or have the necessary permission to use. Commercial permission does not guarantee copyright protection or exclusivity. See the Terms of Service.
Your uploads and generated images are private by default in your account. Your prompts, uploads and outputs are not used to train AI models. See the Privacy Policy for processing and storage details.
