Recraft V3 Image to Image for Photo Edits and Restyles

Last verified: October 19, 2026

Recraft V3 works best here on targeted changes to a picture you already have, such as a different season, a recolored object or a new medium, when the rest of the frame should still read as the original. In Recraft V3 image-to-image, the input image becomes a new image that keeps certain parts of the source and changes others based on the prompt. That is what makes it useful for production work: you keep the shot that was already approved and change only what the brief asks for.

The form has two inputs, Image and Prompt, so every decision about the result sits in your wording. The picture gives the starting point, and the prompt says what should change and what should stay as it is. Each run takes one upload and one instruction, so you can judge one change at a time before you move on to the next.

Explore more image to image

Image To Image Seededit V3 Nano Banana 2 GPT Image 2 Seedream 4.5 Edit Qwen Image 3.0 Qwen Image 3.0 Pro
Capability Snapshot

Recraft V3 Inputs and Output at a Glance

What goes into each edit and what you get back.

Input

Image (JPG, JPEG, PNG or WEBP, up to 10 MB) + a text prompt of up to 2,048 characters

Output

1 image per run

Prompting Tips for Better Edits with Recraft V3

Describe the Change, Not the Whole Scene

In this mode, the prompt is a description of the areas to change. Recraft's guide splits the work this way: the source image provides the composition, colors and subject, and the prompt provides the change. So write what becomes what, for example the sky turned overcast or the walls painted sage green, instead of describing everything the picture already shows. That keeps the instruction clear and leaves room in the Prompt box for the details that matter.

Name What Must Stay the Same

For edits and restyles, Recraft recommends saying what stays. A phrase that keeps the composition, lighting and camera angle holds the frame in place while the rest of the prompt makes the change. The guide warns that without those words the whole image is open to change. When the layout matters, end your prompt with the parts that must stay, such as the pose, the background and the framing.

How to Edit an Image with Recraft V3

Four steps from upload to finished picture.

Recraft V3 or Another Way to Rework an Image

When an upload-and-prompt edit is the right choice, and when another route fits better.

Starting from a picture or from text Your upload is the starting point, and the model keeps certain parts of it while changing others based on your prompt. Text-to-image generation through Recraft's Generate image operation creates a picture from a prompt alone, so every part of the frame comes from your words. Choose the upload route when a photo's layout has to survive the change, and choose text generation when there is no source picture to keep.
Giving a written instruction A prompt is required with every upload, and it is where you state exactly what should change. Remix image in Recraft Studio needs no prompt and gives new versions that keep the overall idea of the image while the details move. Use Recraft V3 here when you know the specific change you want, and use Remix when you like the idea of an image but want loose takes on its details.
One-off edits or batches Each run takes one upload and one prompt, which suits edits you check by eye, one picture at a time. The Recraft API runs the same image-to-image job without a canvas, and Recraft suggests it for batches of product shots and for pipelines. Edit individual pictures here when every result needs a human check, and move to the API when the same change has to go through a large set of images automatically.
Use Recraft V3 here for a specific, written change to a single picture you want to keep. Choose text generation when there is no source image, Remix when you have no instruction to give, and the API when the job is a batch.

Common Problems When Editing Images with Recraft V3

What usually goes wrong with an edit, and what to change before the next run.

The edit changed small details I didn't ask it to touch

Recraft's guide says small curved shapes are where an edit most often changes more than you asked for. A prompt that names only the change gives nothing for those details to hold on to. Add the specific elements you want to protect to the prompt, such as the handle of a mug, the outline of a logo or the subject's pose, and run it again.

The result barely looks like my original photo

The model reads the prompt as a description of the areas to change. A prompt that describes a whole new scene is asking for exactly that. Cut the prompt back to the one change you need, or split a big transformation into separate runs that each handle one change, then generate again.

My image won't upload

The Image field only accepts the file types, file size and minimum dimensions listed in the uploads FAQ. Convert the file to one of the accepted formats, compress it if it's too large, or use a bigger original if it's below the minimum size. Then upload it again. You don't need to change the prompt.

I can't fit my whole prompt in the box

The Prompt box has a character limit, and long prompts often use it up describing what the picture already shows. Delete the description of the existing scene, keep the change and the list of what must stay, and paste in the shorter version before you click Generate.

Frequently Asked Questions

Which image files can I upload to Recraft V3?

The Image field takes JPG, JPEG, PNG and WEBP files up to 10 MB, and the picture must be at least 300×300 pixels. Each edit starts from one uploaded image.

Do I need to write a prompt, or can I just upload an image?

You need both. The Prompt is required alongside the image, because it is how you tell Recraft V3 what to change. One clear instruction is enough for a first run. If the result moves too far from your original, add the parts that must stay the same.

How is the cost of a run shown?

Once you're signed in, the credit cost of the run is shown above the Generate button. It is worked out from what you've set on the form.

How many images do I get per run, and how do I get variations?

Each run returns one image. For alternatives, run the same picture again with a reworded prompt. Recraft describes a variation as a job that keeps the idea of an image and changes its details, so change the details you ask for rather than rewriting the whole instruction.

References

Sources and citations used to support the content provided above.

Updated: 2026-10-19 23:37:04 3 Sources

webflow.recraft.ai

Source Link
https://webflow.recraft.ai/docs

www.recraft.ai

Source Link
https://www.recraft.ai/docs/api-reference/examples

www.recraft.ai

Source Link
https://www.recraft.ai/blog/ai-image-from-another-image