Some edits are slow by hand: a new background, an object taken out, a photo turned into a painting. Attach the photo, say what to change, and an AI image model on a server makes the edited version. You sign in first, and the photo is sent only with your permission.
Change the background, add or remove objects, change the style, restore or enhance, or recolor parts of the picture.
The result is square, landscape (3:2) or portrait (2:3), whichever is closest to your photo.
Ask for the quality you want, and the card shows it. Higher quality costs more credits.
A permission card names the photo before it is sent, unless you already allowed uploads for this tab.
The card shows the estimated credits. They are held during the edit and refunded if it fails.
The edited picture is a new file in Output Files; the photo you attached is not changed.
Sign in with Google or Apple, then add a JPEG, PNG or WEBP photo with the + button.
Say what to change. ToolGPT asks permission to send the photo, then shows a card with your instruction, the quality and the cost.
Press Edit picture, or reply “go”. The edited picture arrives in the chat with a download button.
The editor works from a written instruction. Good requests say what should change and what should stay: “replace the sky with a sunset, keep the people as they are”. It can replace backgrounds, add or take out objects, restyle a photo, restore old pictures and recolor things such as clothes or cars.
Each edit is a new attempt by the AI model, so results can vary. If the first version is not right, change the instruction and ToolGPT shows a new card.
This is one of the few ToolGPT tools that sends your file away. The photo goes to ToolGPT's server and on to an outside image service running an OpenAI model. Before the upload, a permission card names the photo and waits for you: choose Allow once, or Allow for this tab to skip the question for later uploads in the same tab. The assistant cannot give permission for you.
Once you allow it, the photo is uploaded to ToolGPT's server right away, before the cost card appears; it goes on to the image service only when you confirm. You can edit a picture you generated earlier the same way.
Editing needs an account; signed out, you see a Sign in button. Each edit is charged in credits: the card shows an estimate, the credits are held while the model works, and they are refunded if the job fails, for example under the image service's content rules. The final charge is the actual cost. There can also be a daily limit on how many edits an account starts.
An edit usually takes under a minute, and you can keep chatting meanwhile. The result is saved to Output Files, ready for the regular photo tools.
Yes. AI editing needs an account; sign in with Google or Apple, then ask again.
Yes, after you allow it, because the edit is made on a server. Unless you chose Allow for this tab earlier, a permission card names the photo before it leaves your device.
Close to it. The result comes out square, 3:2 landscape or 2:3 portrait, whichever is nearest your photo's shape.
No. Each edit works on one photo; ask again for the next.
Try it with your own file
Open ToolGPT, attach a file, and type what you want done.