Hunyuan 3D 3.1 Create 3D Models from Text or Images

Choose Your Hunyuan 3D Output

Decide whether you need shape, surface appearance, or both before generating.

For an early proportion check or a model you plan to paint yourself, choose Geometry. For a prop or character concept that needs its colors and surface appearance, choose Textured. PBR adds a separate material option; it is not required just to receive a textured model.

Face count is a complexity setting, not a guarantee of accuracy. Increasing it cannot supply a missing back view or correct an ambiguous shape. If the main proportions are wrong, improve the input before requesting a denser mesh. If the shape is useful but the file is heavy, consider a lower target or simplify the downloaded result.

Prepare Your Text or Image Input

Give Hunyuan one clear object to interpret, using either words or a reference image.

Starting from text

Describe the object’s main shape, defining parts, material, and style. Prioritize details that should be visible in the model, rather than camera effects or a surrounding scene. Tencent’s prompt guidance emphasizes a single subject and does not require quality tags such as “4K” or negative-prompt formulas.

The text input accepts up to 1,024 characters. Use that space to resolve ambiguity: decide which parts are connected, what proportions matter, and which material belongs to each major surface. Text mode does not use an accompanying image as a second instruction.

Starting from an image

Upload a JPG, PNG, or WebP of up to 6MB. Keep the entire object visible, with a simple background and enough separation between its parts. A reference that shows the front and a side can reveal more shape information than a flat frontal view, but choose the angle that best explains your particular object.

Use the image generator if you need to develop a reference first, or remove a distracting background from an existing image. Background cleanup can clarify the outline; it cannot reveal an obscured part or tell Hunyuan what the unseen side looks like.

Hunyuan 3D 3.1 vs 3.0

Version changes matter most when they affect the inputs or output modes your workflow needs.

Tencent’s current API documentation distinguishes these options:

On 3D Design AI, Hunyuan accepts text or one image and offers Geometry or Textured output. The official service’s additional views, LowPoly, and Sketch options are not available in this generator.

Lowering the face-count setting is therefore different from selecting a dedicated LowPoly mode. For multi-image generation on this site, explore the supported inputs in the 3D model generator.

Hunyuan 3D or Trellis 2?

Choose between them by the input you have and the settings you want to control.

Hunyuan is a practical choice when you want to begin with words or explicitly request geometry without generated textures. Trellis 2 is worth exploring when you already have an image and want its resolution and texture controls. Its separate shape-resolution and texture-size controls let you adjust those aspects of an image-based asset independently.

Compare results from the same reference, focusing on the outline, part connections, and surfaces that matter to your project.

From Generated Model to Usable Asset

Match your next step to what the model will actually be used for.

Props and character concepts

Use the preview to check the silhouette, proportions, and connections between parts. Inspect arms, handles, legs, and other narrow joins from more than one angle. If two parts have merged or an essential feature is missing, revise the reference or description. Plan any animation topology and rigging separately.

Scene and product visuals

Import the downloaded model into your 3D editor and review its orientation, scale, and materials under your scene lighting. Check seams and areas that were hidden in the reference. A generated product concept can help communicate a design, but it should not be treated as a dimensionally accurate copy for technical approval.

Print preparation

Review the actual mesh for holes, disconnected pieces, and fragile sections. Set the intended physical size before assessing wall thickness or supports. Geometry mode provides an untextured starting point for this preparation.

For the broader input, preview, and download workflow, see the AI 3D model generator guide.