3D GenerationActive
microsoft / Trellis-2
Trellis-2
Convert images into detailed 3D meshes using Microsoft's Trellis-2 model. Supports various resolutions and customization options.
3D GenerationH200
Model ID
Trellis-2
Provider
microsoft
Updated
1779102968
wiro playground—microsoft/Trellis-2
Updated 1779102968
Overview
Trellis-2 is a deep learning model developed by Microsoft designed to convert 2D images into detailed 3D meshes. It supports multiple resolution options and provides advanced parameters for fine-tuning the output.
What you can build
- 3D models for games and animations
- Digital assets for virtual reality experiences
- Product visualizations and architectural renders
Inputs
- Input Image: The source image to be converted into a 3D mesh. Background removal is automatic if no alpha channel is present.
Outputs
- A 3D mesh in GLB format
- PBR texture maps at configurable resolutions
Recommended settings
- Use
1024_cascadefor a balance of speed and quality - Set
seedto 0 for randomized results - Configure
texture_sizebased on desired render quality
Limitations
- Requires compatible hardware for optimal performance
- Output quality depends on input image clarity and detail
- Advanced parameters require technical knowledge
Safety & compliance
- Designed for creative and commercial use
- Follows Microsoft’s AI ethics guidelines
- No explicit safety measures required for standard usage
API quick start
Run Trellis-2 with a single API call.
POST https://api.wiro.ai/v1/Run/microsoft/Trellis-2
{
"inputImage": "https://your-cdn.com/input.png",
"pipeline_type": "1024_cascade",
"seed": 0,
"texture_size": 4096
}