ワークフローに戻る

HY 3D 2.0 MV

Generate 3D models from multiple views using Hunyuan3D 2.0 MV.

ワークフローをダウンロードCloudで試す

Input

The same object seen from behind, same size and framing.

The object seen from the front, alone on a plain white background.

Advanced settings

How closely the shape follows the views.

Default: 7.5

Higher keeps finer detail but is slower.

Default: 256

Leave empty for a new result each run. Enter a number to repeat one.

Sampling steps.

Default: 20

Run

Output
00:00
Example

例を試す

このワークフローについて

The 'HY 3D 2.0 MV' workflow is designed to convert 2D images from multiple views into detailed 3D models using the Hunyuan3D 2.0 MV model. This process leverages advanced AI techniques to interpret and reconstruct the spatial geometry of objects from images taken from different angles. The workflow uses a series of nodes, including KSampler for sampling, CLIPVisionEncode for image encoding, and VAEDecodeHunyuan3D for decoding into 3D space. The Hunyuan3Dv2ConditioningMultiView node is crucial as it integrates multiple image perspectives to enhance the 3D reconstruction accuracy.

This workflow is particularly useful for applications requiring precise 3D model generation from limited visual data, such as virtual reality content creation or digital twin development.

Technically, the workflow begins by loading the Hunyuan3D model and the multi-view images through the ImageOnlyCheckpointLoader and LoadImage nodes, respectively. These images are processed to extract latent features, which are then conditioned using the Hunyuan3Dv2ConditioningMultiView node. The ModelSamplingAuraFlow node refines the model's sampling, ensuring high fidelity in the resulting 3D model.

Finally, the VoxelToMesh and SaveGLB nodes convert the voxel data into a mesh format and save it as a GLB file, making the 3D model ready for use in various applications. This workflow is a powerful tool for anyone needing to generate high-quality 3D models from 2D images efficiently.

よくある質問