Is my photo uploaded?

No — this page has no input image. Creating the picture sends only the prompt you wrote; when the canvas already has content, the assistant takes a thumbnail screenshot to look at it. The ones to tell apart are the 10 in the AI panel (old photo restore, colorize, replace background, ID photo, upscale, beautify, decorate, de-age, plus Split into editable layers and Edit text in image) — they work on your photo, so they upload it, and carry a cloud icon next to their title. As for compression, cropping, cutout and the other tools, they all run in your browser and not a single byte leaves.

Free, with no sign-in first and free credits every day. Every generation uses up credits, including the re-runs you weren't happy with — a cloud call costs us real money, and we're not hiding that. Local tools (cutout, compression, cropping, colour and the like) are unaffected: still no sign-in, still no limit.

Can I use the output commercially?

That depends on the terms of the model provider behind it, not on us, so we're not going to hand you a «yes, go ahead commercially» here — that's a promise we have no standing to make. The sensible approach: treat it as material rather than a finished piece, and check the current terms yourself before anything that matters. What we can promise is the other half: we don't reuse your results.

It doesn't look like what I described. Now what?

Usually the prompt is too short. The model can't see the picture in your head, so separating subject, style, light and framing beats piling on adjectives: «a cat» → «a ginger cat on a wooden windowsill in warm light, shallow depth of field, film grain». Also don't expect accurate lettering or hands — those are the well-known weak spots of this kind of model. If you need text, add it afterwards with the editor's text tool, where it also stays editable.

AI image generation: write a sentence, get a picture, keep editing it

Generating isn't the finish line. The moment the image lands on the canvas, you still have a whole editor.

Use it online now

Most text-to-image tools end at «download»: generate, glance, rewrite the prompt, try again. But what usually happens next is editing — crop off the excess, warm it up a little, add a line of text, combine it with another picture. Here generation sits inside the editor: the result is a layer, and from there you use the same tools as always, with no export-and-reimport in between.

This page doesn't upload your images, because it doesn't need an image at all. Text-to-image's only input is the prompt, and that's all the creation step sends (when the canvas already has content, the assistant takes a thumbnail screenshot to look at it). By contrast, old photo restore, replace background and the rest of the 10 in the AI panel work on your photo, so they upload it — they carry a cloud icon next to their title. Generation itself runs in the cloud — no sign-in, free credits every day.

Key features

The result is a layer, not a file you can only download

Once generated it goes straight onto the canvas, as a new layer by default (“Apply to” also offers “In selection” and “Replace current layer”). Then crop, curves, text, masks, compositing with other layers — whatever the editor has, this picture can use.

Optional transparent background

Tick “Transparent background (stackable)” and the AI is asked for an image with a transparent background, ready to stack on something else.

You can generate into a selection

Draw a box on the canvas first and switch “Apply to” to “In selection”: the result goes into that box — much easier for a local replacement or filling a corner than repainting the whole thing.

Watch progress in the chat

Once sent, follow the progress in the AI assistant chat, and stop it any time if you change your mind.

How to use

1
Open a document first

The generated image has to land somewhere, so create one at the size you want (or open an existing picture).
2
Write the sentence

The more specific the better: subject, style, light, framing. «A cat» and «a ginger cat on a wooden windowsill in warm light, shallow depth of field» are worlds apart.
3
Wait

Once sent, follow the progress in the AI assistant chat and stop it any time. While you wait, you can still work on other layers.
4
Keep editing once it's on the canvas

The image goes straight onto the canvas as a new layer; then crop and adjust as needed. Not happy, delete the layer and go again — but note every generation uses a credit.

FAQ

Is my photo uploaded?: No — this page has no input image. Creating the picture sends only the prompt you wrote; when the canvas already has content, the assistant takes a thumbnail screenshot to look at it. The ones to tell apart are the 10 in the AI panel (old photo restore, colorize, replace background, ID photo, upscale, beautify, decorate, de-age, plus Split into editable layers and Edit text in image) — they work on your photo, so they upload it, and carry a cloud icon next to their title. As for compression, cropping, cutout and the other tools, they all run in your browser and not a single byte leaves.
Is it free?: Free, with no sign-in first and free credits every day. Every generation uses up credits, including the re-runs you weren't happy with — a cloud call costs us real money, and we're not hiding that. Local tools (cutout, compression, cropping, colour and the like) are unaffected: still no sign-in, still no limit.
Can I use the output commercially?: That depends on the terms of the model provider behind it, not on us, so we're not going to hand you a «yes, go ahead commercially» here — that's a promise we have no standing to make. The sensible approach: treat it as material rather than a finished piece, and check the current terms yourself before anything that matters. What we can promise is the other half: we don't reuse your results.
It doesn't look like what I described. Now what?: Usually the prompt is too short. The model can't see the picture in your head, so separating subject, style, light and framing beats piling on adjectives: «a cat» → «a ginger cat on a wooden windowsill in warm light, shallow depth of field, film grain». Also don't expect accurate lettering or hands — those are the well-known weak spots of this kind of model. If you need text, add it afterwards with the editor's text tool, where it also stays editable.

Try AI image generation: write a sentence, get a picture, keep editing it now

No install · no sign-up · local stays local · free AI credits every day

Use it online now