From image to textured 3D
Generate a single 3D asset from scratch and texture it — all in one pipeline, built in a Blank project.
This is a single-asset route. If you want a whole environment decomposed into many assets, see Generate a scene.
Generating geometry is only half the job — an untextured or flat-shaded mesh isn't game-ready. Running it through I2M grounds the asset in a predicted PBR material stack, so what you export drops straight into a render or game pipeline.
Nodes: Generate Image → Generate 3D Model → Generate PBR Textures
When to use this
You're starting from an idea, a prompt, or a single reference image and want one fully textured 3D asset at the end.
Steps
Create a new project, choose Blank, and build the image → 3D route: Generate Image → Generate 3D Model → Generate PBR Textures.
In Generate Image, enter a prompt. (Prefer your own image? Swap in a Load Image node and upload it instead — uploading is free.)
Generate 3D Model turns that image into a mesh. Pick a generator — Hunyuan 3D 3.1 Pro is the default; Tripo H3.1, Meshy 6, Meshy 7 and Meshy 7.1 are also available, at different credit costs.
Generate PBR Textures applies the material stack.
Preview in the 3D viewer, then export the textured asset and its maps.
Variations
Upload your own image — replace Generate Image with a Load Image node. Uploading is free.
Refine the image first — add an Edit Image node before generating the 3D model.
Improve the 3D reconstruction — use a Generate Multiview node to produce multiple views of the object before Generate 3D Model. Multi-view input costs more than a single image, and only Hunyuan 3D 3.1 Pro accepts it.
Set a specific face count — pick a custom polycount on Generate 3D Model. Every generator offers it, but only Hunyuan 3D 3.1 Pro charges extra for it, and on Hunyuan it's single-image only.
Try a different generator — none of them wins on every subject, so re-running the same image through another one is often a faster fix than re-prompting. They're priced differently, so check the cost first — see Choosing a generator.
Tips
The object must fill most of the image. This is the single biggest cause of a failed 3D generation — a subject that's small, distant or lost in a busy frame won't reconstruct.
A clean, well-lit source image with clear shading produces a better mesh and better materials.
Flat-shaded generative images predict poorly. Shading and lighting cues baked into the base colour give I2M something to work from.
Change the generator before you change the image. If the reference image is good and the mesh still isn't, the reconstruction model is the variable worth moving. See Choosing a generator.
See the Nodes reference for what each node costs, and Credits & billing.
Last updated