Skip to main content
Access: Public guide. Modeler preview creation requires approved beta access and sufficient Credits. Other API services follow their current access requirements. Image-to-3D uses a picture to guide the appearance of a generated 3D object. Mint supports image-guided 3D Model generation and reference-image inputs for Modeler. A photograph is a visual reference; unseen surfaces still need to be inferred.

​
Choose the right reference

Use a clear view of one subject with its silhouette visible. Avoid heavy occlusion and distracting backgrounds when possible. If you have multiple views, choose images of the same object with consistent proportions and materials. Tell Mint what to preserve: the overall shape, a distinctive handle, a surface finish, or a particular proportion. Describe ambiguous details rather than expecting a single view to establish hidden geometry.
Only upload references you have permission to use. A generated model is not a measurement-certified reconstruction of the photographed object.

​
Match the image to the output

An image-guided Asset Pack creates individual models from a coordinated plan. It does not return pixel masks or guarantee recovery of every object in a photograph. World generation produces an environment representation rather than an ordinary object GLB.

​
Use reference images through the API

For standard 3D Model generation, send one imageUrl or 2–8 sourceImages, never both. Read the model service guide for the current request shape. The default automatic workflow continues to final generation. Choose review mode when your integration needs an explicit Preview approval step. Modeler uses a different input contract: upload images through the reference-image API and pass the resulting owned referenceImageIds. An image-only Modeler request may omit the prompt. Do not substitute imageUrl for Modeler’s reference-image IDs. Keep the operation ID, wait for completion, then retrieve the returned asset and its files. A source image, a Preview, and a finished model are separate artifacts.

​
Review shape and materials

Rotate the delivered model and compare it with the reference. Check the front, sides, and back, with particular attention to hidden surfaces and thin parts. Check dimensions in your target editor if physical size matters. For a Modeler result, describe a focused change through revision. For the standard generation service, use Preview revision before approving final generation when you need that review step.

​
Common questions

​
Is image to 3D the same as photogrammetry?

No. Image-guided generation creates an object from visual references and can infer surfaces that were not photographed. Do not treat its output as a survey or exact scan. Use the capture guide when you want Mint’s mobile capture workflow.

​
Can I turn a room image into separate furniture models?

Use an Asset Pack brief for the visible furniture you want to create as individual items. Specify an item list and shared style. Use a World when you want the room as an explorable environment instead.

​
Can I download the result for Blender or a web project?

Finished models have a GLB when delivery succeeds. Retrieve supported files from the asset response or manifest. The Blender add-on can import finished Mint GLB models; browser projects can load the returned files with their chosen glTF loader. Continue with text to 3D, 3D file formats, or workflow selection.