Lovable now edits the images in your project.
Since November 10, Lovable can reshape an image, erase parts of it or animate it without leaving the project. For design, what matters is less the tool than the image you feed it.
In one week Lovable's changelog added three image items: an editing toolbar in the Files tab, the same toolbar in Chats, and new image models for the AI features of apps built there.
What follows is what its documentation says, plus one opinion: editing a good image improves a site; generating any image makes it interchangeable.
What Lovable announced
Per Lovable's changelog, Resize and Make video arrived in Files on November 10. Resize changes an image's shape (square, portrait, landscape, widescreen) and extends the scene instead of cropping it. Make video animates the image, from a prompt or with a quick animation of subtle motion.
A day later the same toolbar showed up in Chats, with a difference the docs spell out: in Files each result is a new file next to the original, while in Chats the edit is saved as a new version of the same image. The toolbar also lets you pin points with one change each, erase an area or remove the background.
The image models for your apps
On November 6 the changelog logged something else, separate from the toolbar: two image models for the AI features a published app can offer. GPT Image 2.6 Flare is the faster one, meant for previews and in-app editors. Sunburst is for detailed visuals and precise edits. Both can output transparent backgrounds. Neither is available with EU inference turned on.
This does not change how your site looks. It changes what your app can offer the people using it.
Where it helps design, and where it does not
It helps with the mechanical work: the same photo as a square for a card and as widescreen for the hero, without cutting a subject in half, or your own image minus a leftover background. That saves time and invents nothing.
It does not help if the starting image is generic. A tool that extends a scene with what "should" be around it produces the most likely thing, and the most likely thing is what you have already seen on a thousand sites. My view: use it to adapt your own photos, not to fill gaps with images nobody chose.
Edits also cost credits, like any other build message plus the cost of generating the image. Know what you want first.
What I would do this week
A prompt to start with:
- Upload your own photos or real product captures to Files.
- Ask for the formats you need, one result per file, and discard whatever does not keep the subject the same.
- Check each version at full size: an extended scene often gives itself away at the edges.
- Write an
altthat describes the image you see, not the one you asked for.
Use the image I uploaded (shop-front.jpg) as the base. Make it
widescreen for the hero and square for the cards. Do not change
the subject, the colors or the text on the sign. Save each
format as a separate file and tell me which one you used in
each section.Then measure the result: the Slop Score flags the generic-design signals it can see on the page, from the field on the homepage. And if you wonder how to keep a style of your own when you prompt an AI, see the post on Lovable's AGENTS.md files.
Sources
- Changelog, Lovable Docs, November 20, 2026
- Generate and edit files, Lovable Docs, November 20, 2026
Frequently asked questions
Can Lovable resize an image without cropping it?
Yes. Its documentation says Resize extends the scene to fill the new frame instead of cropping it.
What happens to the original image when I edit it in Lovable?
In Files the result is saved as a new file and the original is unchanged. In Chats it is saved as a new version of the same image.