A photo you already like can be a starting point, not a finished frame. The free image to image tool on CanvaPro takes a reference file and a transformation prompt, then returns a new variation you can preview and download. You do not need an account, and the usual file does not pick up a watermark from the page.
The live uploader accepts a JPG, PNG, GIF, or WEBP up to 8 MB. A reference image and a transformation prompt are both required. You can also click Generate Prompt if you want help drafting the text from the photo. Aspect ratios on this page are Square (1:1, the default), Portrait (9:16), and Landscape (16:9) — a shorter list than on some of the named-model generators. After you pick a model, complete the security check and click Generate Variation. Jobs usually finish within about a minute.
Models on the form
The model list is GPT Image 2 (the default), Nano Banana, Flux, Grok, GPT Image 2.5 Flare, Ideogram, Seedream 5 Lite, and Seedream 5 Pro. If you are comparing models, change only the model between runs so you can see the difference clearly. The output will not look exactly like your reference; the point is a variation guided by the photo and the prompt.
How to transform a photo, step by step
- Upload a JPG, PNG, GIF, or WEBP under 8 MB, and only use a file you have the right to edit.
- Write a transformation prompt, or click Generate Prompt and then edit the text so it says what should change and what should stay.
- Choose Square, Portrait, or Landscape, pick a model, complete the security check, and click Generate Variation once.
- Review the preview and download it if the new still is useful. Keep earlier versions so you can compare models or prompt tweaks.
Prompts that respect the reference
Name the style or change you want, and mention the parts of the photo that should remain — a pose, a product shape, a composition. “Keep the cat and the windowsill; restyle as an impressionist oil painting” is more useful than “make it art.” Creators use this path for look development, product mood variations, and social drafts. Treat every result as synthetic media, and label it when a client or platform needs that context.
A pencil or ink drawing that should look like a photograph has its own walkthrough: turn a drawing into a realistic photo.
If the variation drifts too far
Tighten the prompt, confirm the 8 MB limit, and try a different model. A busy server can add a short wait. If you would rather stay inside Flux Kontext specifically, with a wider aspect list and the same kind of edit, open the Flux Kontext tool and run Image to Image there.
Uploads and prompts travel over HTTPS. Requests are handled to complete generation and are not sold.
Fusing two or three stills into one frame is a different job, handled by the AI image combiner rather than a single-image variation.
After the variation appears
Compare it with the original at the aspect you chose. If the composition is close, change one variable and generate again from the same reference, rather than transforming an already-varied export. Save the file with a note of the model and prompt.
Compare it with the reference at 100 percent, and if the design slipped, tighten the prompt and try the image-to-image tool again from the original still.
Picking a model and a shape
GPT Image 2 is the default, but it is worth trying another model on the same prompt when the first variation is close but not quite right. Ideogram can be a useful second pass if the still includes a short label. Seedream 5 Lite or Pro, Flux, Grok, Nano Banana, and GPT Image 2.5 Flare each have a different feel. Change only the model between those tests so you can actually compare.
Square, Portrait, and Landscape are the only aspect choices on this page. If you need a 4:3 or 3:2 frame, generate in the closest option and crop carefully, or use a generator with a wider aspect list. Set the shape first so a vertical idea is not squeezed into a landscape variation you will have to recrop.
The Generate Prompt button is a shortcut, not a finished brief. Read it, then add what should stay and what should change before you click Generate Variation. A prompt you have actually edited usually respects the reference more faithfully.
When the edit must stay tightly anchored to the original still, edit the image with Flux Kontext is often the more controlled option.
A short checklist
- Upload a JPG, PNG, GIF, or WEBP under 8 MB, plus a transformation prompt — both are required.
- Choose Square, Portrait, or Landscape, then pick a model.
- Click Generate Variation once and wait about a minute for the preview.
- Download the still you like, and keep the original reference.
For a set of variations, keep the same reference and aspect ratio, and change only the prompt or the model. That makes it much easier to choose a winner, and it avoids a folder of stills that no longer share a composition.
When you are ready, upload one photo, write one clear change, and save the preview only after it still feels like the same scene. A single considered variation is easier to use than a long string of loose restyles.