Img aiImg ai
Img ai
Get 50% OFF on Annual Plans→

Generative Fill

Paint the part of a photo that should change, describe what belongs there, and get the same photo back with only that region redrawn.

Upload the photo you want to fix

JPG, PNG or WebP up to 20MB. Both sides need at least 256px. Your image is resized to the fill grid so the result lines up exactly.

0 / 1000

This run costs

10 credits

Your balance

--

Every run is saved to My Creations

What people use it for

Swap the hero product in an approved packshot

The set, the light, and the cloth were signed off; only the bottle has to change for the next SKU. Paint the bottle, name the new container, and the stone surface and cast shadow stay where the photographer put them.

Stage an empty surface in a listing photo

A bare side table or console reads as unfinished. Paint the empty top and describe a lamp, a stack of books, or a bowl that suits the light already in the room.

Change the art inside a frame

Keep the wall, the molding, and the furniture below, and replace only the print. The inner edge of the frame is a natural mask boundary, so the seam has somewhere clean to land.

Clear a stray object from a scene

Object removal is the same operation with a quieter prompt. Paint the coffee cup on the desk and describe the desk surface that should continue underneath it, rather than a new subject.

Rebuild one damaged or dull region

A blown-out window, a scuffed patch of floor, or a dead corner of sky can be repainted from a short prompt while the subject and the framing are left alone.

How it works in three steps

  1. 01

    Upload a photo

    Choose a JPG, PNG, or WebP up to 20MB whose shorter side is at least 256 pixels. Sign in first; upload and generation both run against your account. Before anything is sent, the browser fits the working size to the model's 32-pixel grid and caps the long side at 1440 pixels.

  2. 02

    Paint the area to change

    Brush over every pixel that should be redrawn, including a thin margin past the edge of the object. Painted pixels are repainted and everything outside the stroke is preserved. The mask has to stay inside the frame; the tool does not extend the canvas.

  3. 03

    Describe what belongs there

    Write 3 to 1000 characters naming the object or surface that should appear, its material, and the light it sits in. Describe the fill, not the whole picture. Each generation reserves 10 credits and returns a PNG.

Before and after, as concept examples

Each pair keeps the same scene, camera, and light. One painted region differs between the two frames, because that is the only thing the tool is allowed to touch.

Concept example. These pairs are generated illustrations of the mask-and-prompt job, not measured customer results.

Matte ceramic diffuser bottle on a warm travertine surface beside a folded linen cloth, before the fill
Before
Faceted amber glass jar standing where the diffuser was, on the same travertine surface with the same cloth and shadow, after the fill
After

Concept example

Product swap in a styled still life

Masked region
The diffuser bottle and its cast shadow, with a narrow margin around the silhouette. The linen cloth, the travertine, and the background were left unpainted.
Prompt
a faceted amber glass jar with a matte black lid, the same height as the original, soft window light from the left, warm travertine surface
Mid-century reading corner with an empty walnut side table beside an armchair in late afternoon light, before the fill
Before
The same reading corner with three stacked books and a small brass lamp on the walnut side table, after the fill
After

Concept example

Dressing an empty side table in an interior

Masked region
The empty top of the side table and the air directly above it. The armchair, the rug, the wall, and the window were left unpainted.
Prompt
a stack of three hardback books with a small brass reading lamp on top, late afternoon window light, walnut tabletop
Framed abstract print in muted blue on a plaster wall above a console table, before the fill
Before
The same frame on the same wall now holding a botanical study in warm ochre, after the fill
After

Concept example

Replacing the art inside a wall frame

Masked region
Only the print inside the frame. The frame molding, the plaster wall, and the console table below were left unpainted.
Prompt
a botanical study in warm ochre and cream on matte paper, evenly lit, inside the existing frame

Text inside a painted region is regenerated rather than preserved, and in the recorded provider run it came back garbled. Keep lettering, labels, and logos outside the mask, or plan to set type again after the fill.

What it does and what it refuses

What it does

  • Repaints only the painted region. Unpainted pixels return unchanged.
  • Accepts JPG, PNG, or WebP up to 20MB as the source.
  • Returns a PNG at the working size, up to 1440 pixels on the long side, with each side a multiple of 32.
  • Uses the surrounding photo as context, so a new object inherits the scene's light direction and perspective.
  • Saves each finished result to My Creations on your account.

What it refuses

  • Extend the canvas. The mask must sit inside the uploaded frame; this is not an outpainting tool.
  • Preserve legible text inside the mask. Lettering under the brush is regenerated and typically comes back garbled.
  • Accept sources whose shorter side is under 256 pixels.
  • Produce output above 1440 pixels on the long side. Larger sources are scaled down to the grid.
  • Run anonymously. Upload and generation both require a signed-in account.

Why pick Img ai for this, and when not to

Img ai Generative Fill is a mask-plus-prompt inpainting tool. You paint a region of an existing photo, write what should appear there, and the model redraws only the pixels under your brush while every unpainted pixel is copied through unchanged. That is a different operation from generating a whole image from a prompt, where the model owns every pixel and nothing is anchored to the photo you started with.

The trade is control for scope. Because the unpainted photo is fixed, a new object has to fit the light, perspective, and surface already in the frame, and the parts you did not paint cannot drift. The cost is that the tool only solves in-frame problems. It will not widen a crop, restyle a whole room, or move the camera; for those jobs a full-image model is the right pick.

On Img ai the whole edit is priced and bounded before it runs. A generation reserves 10 credits, accepts one JPG, PNG, or WebP up to 20MB, and returns one PNG at up to 1440 pixels on the long side. There is no layer stack to learn and no desktop install; the mask editor runs in the browser and the result lands in My Creations.

Img ai vs Adobe Firefly Generative Fill vs Canva Magic Edit

Competitor cells state only what is publicly documented and stable. Where a fact was not checked for this page, the cell says so instead of guessing.

AxisImg ai Generative FillAdobe Firefly Generative FillCanva Magic Edit
Where it runsIn the browser at imgai.pro, no install.In the browser at firefly.adobe.com and inside Photoshop.Inside the Canva editor in the browser and in the Canva apps.
Cost of a single edit10 credits per generation, shown on the button before you run it.Draws on Adobe generative credits; the per-edit price is not verified here.Not verified here
Subscription requiredAny Img ai plan that includes credits covers it. There is no separate tier for this tool.Not verified hereNot verified here
Maximum output size1440 pixels on the long side, each side a multiple of 32.Not verified hereNot verified here
File types inJPG, PNG, or WebP up to 20MB, shorter side at least 256 pixels.Not verified hereNot verified here
Output formatPNG.Not verified hereThe edit lives in a Canva design, which exports as PNG, JPG, or PDF.
Results saved to an account historyYes. Every finished fill appears in My Creations.Not verified hereYes. The edit stays in the saved Canva design.
What it refuses to doExtend the canvas, keep text legible inside the mask, accept sources under 256 pixels, or run without sign-in.Not verified hereNot verified here

Pricing and output

Every generation reserves 10 credits on your Img ai account, and the button shows that number before you press it. The quote is fixed per generation; it does not change with image size or prompt length.

The output is a PNG at the working size the browser computed on upload: the long side capped at 1440 pixels and each side rounded down to a multiple of 32. A 1731 by 909 source, for example, returns 1440 by 736.

Finished fills are saved to My Creations, so a result can be opened and downloaded again later.

See credit plansOpen My Creations

Frequently asked questions

Does Generative Fill change pixels outside the mask?

No. The model receives the source and a mask; white painted pixels are redrawn and everything else is passed through. If a result looks different outside your stroke, check whether the brush spilled past the edge of the object.

Why did my result come back smaller than the upload?

The provider works on a 32-pixel grid with a 1440-pixel cap on the long side. The browser resizes to that grid before upload so the result lines up with the source pixel for pixel; a 3000-pixel-wide photo comes back at 1440.

Can I remove an object instead of adding one?

Yes. Paint the object and describe the background that should continue behind it, for example 'bare oak desk surface, same grain and light'. A prompt that names a surface steers the fill away from inventing a new subject.

Can I fill outside the edges of the photo?

No. The mask must sit inside the uploaded frame. Extending the canvas is outpainting, which this tool does not do.

Why is text in the filled region unreadable?

Anything under the brush is regenerated, including type, and the model is weak at drawing letters. Mask around labels and logos, or add the type back in a design tool afterwards.

Do I need an account?

Yes. Upload and generation both require a signed-in Img ai account, and each generation reserves 10 credits from that account. The cost is shown on the button before you run.