By skyforbes 
For a long time, I've been frustrated with the current state of AI image editing. It's either powerful but simplistic editors, or intelligent but imprecise AI tools where you spend more time crafting perfect prompts than actually creating.
So I developed a tool that's different: gemini-image-edit.com
The core idea is simple: since we can directly show images to AI, why rely solely on language to describe visual changes?
You can think of it as a hybrid. It's a visual canvas (based on Fabric.js) where you can draw, but all editing is done by AI.
Here's what makes it unique (and why you might want to try it):
Editing by drawing, not describing: this is the most important point. Want to change your shirt? Draw a box around it and type "leather jacket." Want to remove an object? Draw whatever you want with the brush. AI understands the spatial context based on your markers, generates edits, and automatically removes your markers, leaving only the final image. It's intuitive and easy to use, solving problems like "left side? No, my other left side" in plain text hints.
Major Upgrade: We've rebuilt the front end using Gemini 3, significantly improving UI/UX stability. More importantly, we've integrated Nanobanana Pro. While the operating cost is higher, the results are significant, especially when handling Chinese text and context, showing a marked improvement over previous versions. The quality improvement is obvious.
No Hints Needed: Don't know what to type? We have a huge template library. Find your favorite image style, use its hints with one click, or even search for hints by uploading an image. Truly "what you see is what you get."
Unlimited Canvas, Multi-Image Workflow: Work on multiple images and concepts on a single, massive canvas. Say goodbye to cumbersome tabs.
Precise Tools: Change the color and thickness of boxes, circles, and brushes. Use different colors to mark different types of edits on complex images.
Current Status and Free Credits:
We are currently in public beta. Canvas data is temporarily stored locally in your browser.
To enhance user experience, we offer 500 free credits to every new user, plus 100 credits for daily logins.
These credits are roughly enough to edit 10 Nanobanana images or 2 Nanobanana Pro images.
Design Philosophy:
Most products are either pure AI generators or add AI functionality to traditional editors. What I wanted was a native AI editor built from scratch. Manual editing, while precise, is time-consuming and laborious; AI is powerful, but not precise enough. Why not develop a tool that is both efficient and intelligent? That's our goal.
Try it out and let me know what you think:
There's a built-in feedback button here. I'm still in the early stages of development, so your feedback is invaluable. What features are great? What features have issues? What features are missing?
Hopefully, you'll find it as useful (and fun) as I do.