Edit images with exactly 1.0 edit

Editing is a dedicated mode for changing an existing image. It's powered by the exactly 1.0 edit model (e.edit) and lives in its own tab in the prompt bar.

You can edit any image — one you uploaded from your device or one you generated earlier on the platform. You describe the change in plain text. Optionally, you can also draw rectangular masks to point at specific regions of the image and give each region its own instruction. Both work together — you can have a general instruction for the whole image alongside per-mask instructions for specific areas.

Where to find it

In the prompt bar, switch from the Generating tab to the Editing tab. Drop the image you want to edit into the editing slot; that becomes the source for the edit.

To edit an image you already have in your project, switch to the Editing tab, hover over the image, and click Remix — the image is added to the editing slot automatically.

Editing from the prompt bar

If you only need a text-based edit (no masks), you can do it directly in the small prompt bar:

1. Open the Editing tab

2. Add the image to edit

3. Write your instruction (for example: "make it evening" or "remove the bicycle on the left")

4. Choose the number of variations, resolution, and aspect ratio

5. Click Generate

The aspect ratio of the result can differ from the source image.

Editing in the full dialog (with masks)

To draw masks you need to open the full editing dialog. There are three ways in:

• If the prompt is empty, the prompt bar shows a Draw a mask button — click it

• In the prompt bar dialog, click the expand to full screen icon in the top right

• Click the editing reference image (the first reference image is set to Editing by default) — at the bottom you'll see a Draw a mask button

The dialog shows the source image on a large canvas with the prompt on the right.

Drawing masks

• You can draw up to 3 masks per image

• Masks are rectangular only

• Masks can overlap, but the result quality is not guaranteed in overlap zones

• Masks are optional — you can edit with text alone

When you draw a mask, a section for that mask appears in the prompt where you can write the instruction for that specific region. For example, on a mask covering one flower pot you can write "replace the pink tulip with a red one"; on a separate mask you can write "remove the extra tree"; on a third mask "replace the dog with a cat".

You can keep writing in the general (image-level) part of the prompt at the same time — for example, "change the evening lighting to morning sunrise" applies to the whole image.

Keep in mind that a mask tells the model where to apply the explicit change you described, but the whole image is regenerated on every edit. That means indirect effects can spread beyond the mask. For example, if you draw a mask on the sky and ask for a sun there, the sun will appear in that spot — but the sunlight will also reach the other objects in the image, shadows will shift, and the overall colour balance can change. The mask controls the location of the direct change; the surrounding image adapts to it.

To reference an existing mask from the general prompt, type / and pick the mask from the suggestions. This also works in the small prompt bar once masks have been drawn in the dialog.

Using additional reference images

You can attach additional reference images to your edit, just like in regular generation. Reference them in the prompt with @-mentions.

This is useful when you want to bring an element from one image into another. For example, you have an image of an office and a separate image of a desk lamp:

1. Draw a mask on the office image where the lamp should go

2. In that mask's instruction write: place the lamp from @image-1 in this area

The reference images work inside both per-mask instructions and the general prompt.

Cost and output

• Cost depends on the chosen resolution and starts from 1 credit per edit. See the pricing page for the current rates.

• You can choose the number of variations (same as in regular generation) — the total cost scales with the number of variations

• Resolution and aspect ratio are selectable; aspect ratio can differ from the source image