Meshi AI runs entirely in the browser, so a locked-down school laptop or a managed Chromebook can open it and start modelling within the same minute. Nothing is installed, and no student email is ever collected.

Put the URL on the board and let the class open it. Meshi AI loads in Chrome, Edge, Safari and Firefox on a managed device, so nothing has to clear the software allowlist your IT team guards. Students hand over no email and register nothing, which keeps the lesson out of the district data queue.
Each student types the object the unit calls for, or uploads a photo or a scanned sketch from their notebook. Text to 3D handles things nobody can photograph, like a Bronze Age cooking pot. Image to 3D handles anything they can shoot or draw first. About sixty seconds later a textured mesh appears.
The browser viewer lets students orbit the mesh and argue about what it got wrong before a single file is downloaded. Export STL for the lab printer, GLB for a class web page or headset, OBJ or FBX to keep working in Blender. Geometry arrives manifold and watertight, so the slicer will not fight you.
Classroom software usually dies on three things: installs, permission requests and licence seats. All three are gone here, which leaves the period free for the actual subject matter instead of troubleshooting.
No MSI package, no extension, no admin rights, no lab image rebuild. Chromebooks, iPads and eight-year-old library desktops all get the same tool, because the heavy lifting runs on cloud GPUs and the browser only draws the result. Substitute teachers can run the lesson without a walkthrough.
There is no roster to upload, no student email to gather and no per-seat licence to reconcile at the end of term. Students open a public page, make a model and download a file. That removes the consent-form conversation that usually stalls a new classroom tool for a full semester.
Biology classes model a heart chamber or a cell organelle. History classes rebuild an amphora, a coin or an arrowhead from a museum photo. Chemistry gets tangible molecular props, and design and engineering get brackets, gears and phone stands that actually have to fit something.
The first prompt is almost always too vague, and that failure is the lesson. Students learn to specify shape, material, proportion and detail, then compare two attempts side by side. Writing a precise spec and defending why version three beats version one is transferable far past 3D.
STL exports are manifold and watertight, so Cura, Bambu Studio, OrcaSlicer, PrusaSlicer and Lychee accept them without a repair pass. Set scale in the slicer, check wall thickness on thin features, and a palm-sized student model prints inside a normal class block.
The free tier gives three 3D generations per hour, which maps neatly onto teams rather than individuals. Put students in threes, give each team one generation and one revision, and spend the rest of the period on critique, measurement and slicing. Scarcity forces better prompts.
Open a link, describe an object, download an STL. That is the entire setup for a 3D lesson.
Try Free →Browse what other makers and teachers generated, then hand your students a reference target worth beating.
Open Meshi AI on the classroom projector, type a prompt from today's unit, and have a printable mesh on screen before the bell rings.
Try Free →