Hi everyone! 
I’m an independent developer from New Zealand, and I wanted to share an update on a project I’ve been building called ForgeUI Studio.
The goal is to make embedded touchscreen development easier by combining a browser-based visual UI editor, LVGL-based firmware generation and AI-assisted interface design.
What’s working so far?
ForgeUI Studio lets developers create and edit touchscreen interfaces visually, position widgets on a canvas, configure supported properties and work towards firmware for supported ESP32 display hardware.
My main development target has been the ESP32-P4, including the Waveshare 7-inch touchscreen.
Meet Larry GPU — Our New AI Teammate
Recently, I integrated a locally hosted NVIDIA RTX 4080 running an experimental AI assistant called Larry GPU.
Larry can take a natural-language prompt such as:
“Build a premium marine dashboard with engine temperatures, navigation, battery monitoring and controls.”
He generates an editable UI proposal using ForgeUI’s certified widgets.
Studio validates the proposal, and the developer can open it in the visual editor to move, resize, delete or refine the components.
I’ve been testing him with:
-
Marine navigation dashboards
-
Coffee-machine interfaces
-
Smart-home and hotel-room controllers
-
Aircraft cockpit concepts
-
Retail checkout interfaces
These are real Studio-generated UI drafts, not finished hardware-connected applications.
Still Experimental — And Improving
Larry GPU is very new, and we’re actively improving his layout quality and widget support.
Sometimes he gets spacing wrong or overlaps components. Rather than expecting perfect AI output, the idea is to give developers a useful starting point they can refine.
Studio currently has 81 registered widget types, while Larry’s certified AI composition toolbox exposes 21. We’re investigating how to safely expand that selection.
The GPU inference runs on my own RTX 4080 and communicates with Studio through an authenticated private connection.
I’d Love Some LVGL Developer Feedback
I’m particularly interested in what experienced LVGL developers think about this approach.
Would a visual editor combined with locally hosted AI-generated layouts be useful in your workflow?
Which LVGL widgets or ESP32 display features would you most want supported?
You can see ForgeUI Studio and real examples of Larry’s work here:
Still plenty to build, but it’s been exciting seeing this come together.
Constructive feedback, technical questions and suggestions are very welcome.
Cheers from New Zealand! 