Why it’s interesting
Tavlo AI Lab feels like a row of small AI attractions rather than a single-purpose tool. Nine demos cover live vision, an on-device chatbot, text processing, speech, Pictionary, and doodle recognition. The most immediate contrast comes from reusing the same webcam feed: one demo turns it into a warm-to-ink depth map, another repaints it in the style of famous art, and a vision-language model can describe what it sees.
How to try it
Open the official lab and begin with Pictionary or Creative AI if you do not want to grant device permissions. Draw the prompted object or a digit and watch the model respond locally. For the live vision demos, choose Depth Window, The Narrator, or Style Transfer and allow camera access when the browser asks. Speech and audio experiments require microphone permission. Initial model loading can take longer than the interactions that follow.
Requirements and sources
The lab is free to open in a modern browser and relies on WebGPU for several demos. Tavlo states that camera, microphone, and text input are processed on the device and are not uploaded, recorded, or stored. Some demos therefore need camera or microphone access, while the drawing and text demos offer a no-permission route. Project identity, privacy claims, access details, and the nine-demo count were checked against the official Tavlo AI Lab page and Tavlo site.