In addition to having AI design the entire wireframe (which results in a lot of output that isn’t really relevant or useful to the user’s specific needs), Figma should implement a version of its AI tool that improves/beautifies user-created basic wireframes (eg. takes the text/button names/general layout that someone has already created and adds colors, changes typography, moves things around so they’re visually more pleasing, etc). This would allow users who have a general vision of what they want their product to look like to save time creating more detailed wireframes so they can more easily transition to the prototyping phase.
Question
Figma AI beta
I think figma files can be too big to fit in a context window of an LLM.
Largest context windows of 1m tokens are 3-4Mb and I have .fig files size of several hundreds of MBs.
Currently solutions focus on export / import to and from vibecoding tools where LLMs can digest the distilled code more easily simply because it’s smaller in token size + because code+rendered layout is available in way larger quantities as training data than proprietary figma files + their rendered output.
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.