Image to video turns one still into a short clip.

Open
Image to image

Change one thing in a photo you upload

Upload a photo, name the one change, and generate a new still. The button shows the credits for the model you picked.

Model

Size

2 credits

Outputs

Write a prompt.

Pale wooden chair in a hall with shoes, a tote, and a cable beside it
Remove the shoes, the tote, and the cable. Keep the chair.
The same chair after the shoes, tote, and cable are removed

This example does not fill the upload.

One photograph, one instruction

The upload is the subject. The sentence is the change. A background can go plain, a distraction can leave the frame, a color can shift, and the object you brought should still be identifiable. The form accepts JPEG, PNG, and WEBP up to 10MB. It does not accept a prompt with no file, and it does not offer a resolution above the model's listed output.

Three prices, printed before the run

Each model prints its own credit cost, including the resolution it lists. Changing the ratio does not change that number. Changing the model does, and the button updates before anything is sent. There is no second charge after the still appears.

The result can be the next source

Keep editing takes the finished still and uses it as the next upload. That is a new run. It spends the selected model's credits again. Downloading the file and uploading it yourself is the same kind of run. A video file has no Keep editing control, because this page returns a still.

Failure returns the credits

If the provider rejects the file, times out, or returns nothing usable, the credits for that run come back. You can change the sentence or the model and start again. The original upload is deleted within an hour after the run finishes. It is sent only to make that edit.

Where it fits

One photo

A source you own

The edit starts from the picture you upload, not from a blank frame.

One change

Name the edit

Write the change in a sentence. The model follows that line.

Priced first

See the cost

The button matches the model and the settings before you generate.

Another pass

Edit the result

A later pass is a new run on the still you already made.

What image to image does

Image to image starts with a file you upload and ends with a still. The prompt is an instruction about that file, not a description of a scene that does not exist yet. Replace a background, clear clutter, shift a color, or restore a faded print, and the person or product should remain recognizable. The page does not accept a blank canvas. Models that list 2K or 4K show that tier's credits on the button.

Each model on the form prints its own credit cost. Changing the ratio does not change that number. Changing the model does, and the button updates before anything is sent.

Why the still stays the same subject

A useful edit is a small disagreement with the photograph, not a new photograph. The mug should still be that mug. The coat should still be that coat. Image to image is the page for that kind of instruction. The model menu exists because the same sentence can be worth 2, 4, or 8 credits depending on which priced model you trust with the file, and the button says which one you picked.

Ratios change the frame. They do not change the credit price of the selected model. If you need motion instead of a still, image to video is a different page, and its button shows the credits for that model, resolution, and length.

How an image-to-image run works

Choose a model. Upload a JPEG, PNG, or WEBP of 10MB or less. Write the one change. Pick a ratio if the default frame is wrong. Read the credits on the button. Generate. The stage keeps the source next to the result, and Keep editing can send that result back through the same form, which spends the model's credits again.

  1. 1

    Upload

    Add the photo. This page does not start from text alone.

  2. 2

    Write the change

    One sentence for the part that should change.

  3. 3

    Read the button

    The credit count is the model you selected.

  4. 4

    Generate

    Download the still, or run again. A failed run returns the spend.

For a catalog, a portrait, or a cleanup

Shops use this page when the product is already photographed and the ground, the clutter, or the color is the problem. Portrait work stays inside the same rule: the person you uploaded is the person in the result. Restoration is the same form, with a sentence about the fade or the scratch, not a request for a different face.

A picture that starts from words alone is on the text to image page, and the button there shows that model's credits before it runs. A short clip of a still is image to video.

Plans and credit packs

Lite $17 / Plus $33 / Pro $72 — full list on pricing

Questions

Will image to image run if I only type a sentence?

No. Every model on this page requires at least one JPEG, PNG, or WEBP, at or under 10MB. A sentence with no file does not start.

What is the photo limit on Flux 3 Klein and GPT Image 2?

Four. Nano Banana Edit, Nano Banana 2, and Nano Banana 2 Lite take up to 10. Nano Banana Pro takes up to 8.

Does a new ratio change what image to image charges?

No. The ratio changes the frame. The credits change when you change the model or a priced resolution, and the button updates before the run.

Can Keep editing reuse the still on this page?

Yes. That is a new run and spends the selected model's credits again. This page returns a still, so there is no clip control here.

Related tools