Send a query string parameter q with a sentence to describe a scene. For example:
/?q=An orc tavern with barrels and barmaids and a buring fire in the middle
It will return a JSON array of objects with model paths and 3D coordinates. May take ~30 secs.
Using Shap-E, we can also generate 3D models from text. To do this, just pretend they already exist and you're downloading them as files from "/models/[text here].[ext]". Ext can either be "ply" (original model) or "glb" (converted to GLB with orientation fixed). For example:
It may take a few seconds to load the first time, but after that the models are cached and should load instantly.