Hero image of the tripo-h3.1 image-to-3d detail page: warm-lit negative space on the left; on the right, glass filaments woven into a high-density polygonal sculpture with iridescent cool-toned accents, its name embedded in the glass surface and glowing.

Tripo

Tripo H3.1

tripo-h3.1/image-to-3d

Tripo's high-detail 3D generation model, supporting the generation of textured 3D assets from text prompts, a single image, or multiple views, with adjustable polygon count and texture quality and PBR material output, suitable for asset creation such as game props and product visualization

Try tripo-h3.1/image-to-3d

Input

Reference image

Upload one photo of a single object on a clean background; JPG, PNG and WEBP only.

Output

Fill in the parameters and run — results show up here.

Pricing

tripo-h3.1/image-to-3d per model

Specifications standard detailed
PBR Materials Off · Texture Off $0.290 $0.290
PBR Materials Off · Texture On $0.430 $0.580
PBR Materials On $0.430 $0.580

+ $0.290 finer geometry

Positioning Overview

For developers and design teams that need to quickly generate 3D assets from a single photo. Just provide one HTTPS object image with a clean background, and you get a 3D model with geometry and optional textures, making it suitable for quickly converting real-object photos into digital assets that can be viewed, rendered, or further edited.

Core Highlights

  • Generate a 3D model from a single image: provide a link to an object photo with a clean background to start generation, suitable for quickly digitizing real objects or design drawings.
  • Geometry and texture controlled separately: you can disable both texture and PBR to get only a geometry mesh, or set quality levels for geometry and texture separately depending on the intended use.
  • Face-count limit configurable: you can specify a target face count between 1,000 and 1.5 million triangles (2 million for detailed geometry), making it easy to match different downstream pipelines with different precision requirements.
  • Reproducible results: random seeds can be set for both geometry and texture, and together with fixed parameters this makes repeated generation and result comparison easier.

Capability Description

Image to 3D mesh: input an HTTPS link to a single object photo and output a 3D model; when both texture and PBR are disabled, only a pure geometry mesh is produced, which is suitable for defining the shape first and processing later.

Face count target control: the face limit (face_limit) can be set to a target face count between 1,000 and 1.5 million triangles (2 million for detailed geometry); if left empty, the model decides adaptively, making it easy to match precision requirements for rendering or lightweight scenarios.

Texture and PBR switches: texture is enabled by default, and PBR materials are generated by default; enabling PBR forces texture generation, and only when both texture and PBR are disabled is geometry alone retained, so choose according to downstream use.

Quality level switching: geometry quality and texture quality can each be set to standard or detailed (standard / detailed); the detailed level brings finer mesh details and higher-resolution textures respectively.

Orientation and texture alignment: selecting "align to input image" (align_image) for orientation automatically rotates the model to match the input image; texture alignment (texture_alignment) can be set to fit the original image or fit the geometry.

Applicable Scenarios

  • Converting real-object photos into digital assets: e-commerce or design teams use handheld-object photos to generate 3D models and place them on product display pages or in 3D viewers for users to rotate and inspect.
  • Rapid 3D-ization of concept designs: concept designers convert sketches or reference images into 3D meshes and use them as a starting point for proportions and volumes in modeling software for further refinement.
  • 3D foundation for marketing materials: marketing teams generate textured models from product images and import them into rendering or animation workflows to create 3D product assets for promotional images.
  • Pre-generating lightweight assets: developers set a lower face count and standard textures to batch-generate 3D display assets suitable for mobile or real-time scenarios.

Limitations and Selection Advice

The input accepts only an HTTPS image link to a single object with a clean background; images with cluttered backgrounds or multiple objects in the same frame may not produce ideal geometry. The face count target is limited to between 1,000 and 1.5 million triangles (2 million for detailed geometry), and specifying a low face count loses detail; when texture and PBR are disabled, the output is untextured geometry. Texture alignment and orientation settings affect the final appearance and need to be chosen according to the input image; the generated result is not guaranteed to match the photo point by point.

  • New model tripo-h3.1/multiview-to-3d is live: Tripo H3.1 Multiview to 3D
  • New model tripo-h3.1/text-to-3d is live: Tripo H3.1 Text to 3D
  • New model tripo-h3.1/image-to-3d is live: Tripo H3.1 Image to 3D

Frequently asked questions

View all FAQs →
  • What is the face count upper limit? Can it reach 2 million?

    Yes, the face_limit upper bound is 2 million triangles, but 2 million is only actually supported when geometry_quality is set to detailed; under standard, it is recommended to set it within 1.5 million. If face_limit is not passed, the model adaptively determines the face count.

  • How do I get just a geometry without textures first?

    Set both texture and pbr to false. When pbr is true, textures are forcibly generated. If you need textures but not a PBR material, set texture to true and pbr to false.

  • Should high-density geometry have its own separate parameter?

    Setting geometry_quality to detailed adds finer mesh detail and is also the prerequisite for the polygon count to reach 2 million; if you need more refined textures, set texture_quality to detailed. Both default to standard, so enable them only as required by the close-up needs of the target asset.

  • How can I reproduce generation results so it's easy to compare different parameters?

    Geometry and textures each have their own fixed seed: model_seed controls geometry, texture_seed controls textures, and text-to-3D additionally has image_seed to control the concept image. When comparing parameter settings, first fix the seeds and change only one parameter, so you can tell which item the difference comes from.

Related Models

View all models →