Show HN: I built a web UI for Gemini 2.5 Flash to solve AI image consistency
I wanted to share a project I've been working on, Flash Image.
Like many here, I've been fascinated by AI image generation but consistently frustrated with the 're-roll' loop. Getting consistent characters or making small tweaks without the whole image changing was a huge pain point for me.
When I saw the examples from Google's new Gemini 2.5 Flash model and its potential for multi-turn, conversational editing, I wanted a simple way to test it without code. So, I built this interface on top of it.
The idea is to move beyond a single prompt. You can start with an image and then have a back-and-forth chat to refine it – "make the sky darker," "change his shirt to red," etc. – while trying to preserve the core subject.
It's still early, and not always perfect, but I think this conversational approach is a big step towards more intuitive and controllable AI image generation. I'd love to get your feedback and see what you create.
No comments yet