Upload a reference image, tell the AI what should stay and what should change, then create a new variation without rebuilding the image from scratch.
Reference image
Prompt:Change only the ceramic lamp base color from matte coral red to rich cobalt blue. Preserve the exact lamp shape, circular cutout, cream linen shade, pale oak table, warm-white wall, camera angle, crop, lighting, materials, and shadows. Do not alter any other object or add text, logos, or a watermark.
Start with a photo or sketch, then describe the edit you want. Keep the parts that matter and change the rest.
Turn a selfie, portrait, character, or product photo into a new look without losing the subject. It is useful for professional headshots, outfit changes, editorial portraits, and character restyles.
Example: Turn a casual apartment selfie into a studio business portrait while keeping the same person.
Swap the background while keeping the subject and framing in place. Use it as an AI background changer for product photos, vehicles, portraits, or room scenes.
Example: Move a yellow studio bicycle onto a sunlit European street without changing the bicycle.
Turn a rough sketch into a finished image while keeping its layout and perspective. This works well for interior renders, architectural concepts, and early product ideas.
Example: Render a pencil sketch of an arched reading nook as a warm, photoreal Japandi interior.
Clean up a photo by removing one unwanted item or replacing it with something else. The generator rebuilds the area behind the object so the edit fits the original scene.
Example: Remove a red floor lamp and restore the wall and wooden floor behind it.
Make product color and finish variations from one reference photo. The shape, materials, lighting, and composition can stay consistent across ecommerce and catalog images.
Example: Change a ceramic lamp base from coral red to cobalt blue and leave everything else unchanged.
Give each reference a clear job: one can define the product, while another supplies the color, ingredients, lighting, or overall mood. This makes multi-reference generation easier to control.
Example: Combine an amber skincare bottle with a peach-inspired visual reference to create a product campaign image.
Turn one reference image into a new result in three steps.
Choose a clear JPG, PNG, WebP, or supported image file. Add more references only when each one has a distinct purpose.
Select a model and aspect ratio, then write the preservation rules, requested edit, style, setting, lighting, and exclusions.
Review the current quality, credit total, and visibility setting before generating. Inspect the result and change one instruction for the next variation.
Structure every prompt around three jobs—what to keep, what to change, and what to avoid—so the generator preserves the right details and makes only the requested edit.
Name the face, product silhouette, pose, room geometry, camera angle, palette, or other details that must remain recognizable.
Describe the new background, materials, wardrobe, visual medium, lighting direction, color palette, and level of realism.
List likely failure cases for this image, such as altered logos, extra objects, face reshaping, moved furniture, text, or a watermark.
The usual alternative is moving files and prompts between separate model sites, editors, and download folders. Promptsref keeps reference-led generation and follow-up image work in one browser workflow.
Choose among supported reference-image models in the same generator instead of rebuilding the job in a separate tool.
Available aspect ratios and resolutions, including options up to 4K on supported models, appear before generation.
Save the generated image without a visible watermark, then continue with companion image and video tools when needed.
An image-to-image model creates a new file, but the source material and visible identities may still carry ownership, privacy, trademark, or consent obligations.
Use your own images or material you are allowed to edit. Get appropriate permission before transforming identifiable people or client assets.
Check the visibility control before uploading private material, and avoid sensitive images you would not be comfortable processing under the selected setting.
AI can change faces, product geometry, hands, logos, and text. Review the final file before publishing, selling, or sending it to a client.
Model names matter less than the controls your job needs. Check how many references the model accepts, its available ratios and resolutions, and the live credit total before generating.
| Model | Max references | Resolution | Supported aspect ratios |
|---|---|---|---|
| GPT Image 2 | 16 | 1K / 2K / 4K | Auto · 1:1 · 2:3 · 3:2 · 3:4 · 4:3 · 9:16 · 16:9 |
| Nano Banana 2 / Pro | 8 | 1K / 2K / 4K | Auto · 1:1 · 2:3 · 3:2 · 3:4 · 4:3 · 4:5 · 5:4 · 9:16 · 16:9 · 21:9 |
| Nano Banana 2 Lite | 10 | 1K | Auto · 1:1 · 1:4 · 1:8 · 2:3 · 3:2 · 3:4 · 4:1 · 4:3 · 4:5 · 5:4 · 8:1 · 9:16 · 16:9 · 21:9 |
| Seedream 5.0 Pro | 10 | 1K / 2K | 1:1 · 4:3 · 3:4 · 16:9 · 9:16 · 3:2 · 2:3 |
| Grok Imagine | 1 | — | 1:1 · 2:3 · 3:2 · 9:16 · 16:9 |
Want a different look? Each generator opens with its own examples, prompt tips, and ready-to-use settings.
AI imageUpload one clear photo and turn it into a polished professional headshot. Keep your recognizable features while changing the outfit, background, lighting, and crop for LinkedIn, a résumé, or a company profile.
AI imageUpload a product photo or describe what you sell. Create clean product shots, lifestyle scenes, and ad images for your online store.
AI imageCreate original game character art from a text prompt or reference image. Choose an AI model and aspect ratio, then generate heroes, NPCs, enemies, portraits, and full-body concepts online.
AI imageTurn a prompt or reference image into artwork with cross-hatching, carved marks, and the texture of a traditional print.
AI imageStart with a species and a few features you can picture clearly. Add an outfit, pose, and setting—or upload a reference image—and keep refining the character from there.
AI imageUpload a clear portrait or start with a prompt. Create a recognizable Gigachad-style image with a sharper jawline, sculpted cheekbones, confident expression, and the dramatic lighting that makes the meme instantly readable.
AI imageUpload a clear palm photo to map your life, heart, head, and fate lines. Create a personalized visual palmistry report with labeled details—for entertainment and self-reflection only.
AI imageStart with two clear photos: one of you and one celebrity reference. Load a complete example, replace the sample images, and create a believable backstage, premiere, or stadium selfie without manual compositing.
AI imageTurn a video title, topic, or hook into a 16:9 thumbnail concept. Describe the scene, add a face or visual reference when useful, and compare multiple AI image models without rebuilding your prompt.
It creates a new image from an uploaded reference and a written instruction. The reference supplies visual information such as subject, layout, pose, colors, or style, while the prompt explains what to preserve and change.
Text-to-image starts from a written description alone. Image-to-image adds visual context, making it more useful when you already have a subject, composition, product, character, or style you want the new result to follow.
Yes. A short prompt tells the model how to use the reference. For a controlled result, state what must stay recognizable, what should change, and what the output should avoid.
The limit depends on the selected model. The generator displays the current image limit for that model; some accept one reference while others support several.
Use a sharp, well-lit image with a clear subject and visible defining features. If composition matters, start with framing close to the result you want.
A reference can improve consistency, but AI may still change small facial, product, clothing, text, or design details. Review each result carefully when exact accuracy matters.
Options depend on the selected model. Supported models offer common portrait, square, and landscape ratios, with resolutions up to 4K where available.
The cost depends on the model, resolution, and number of outputs. The current credit total appears before you generate, and failed generations are automatically reviewed for a refund.
Yes. Generated images can be downloaded without a visible watermark.
Upload one clear image, set the creative boundaries, and generate a controlled variation.
