ComfyUI Extension: hitems3D-comfyUI
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.
ComfyUI custom node package for Hitem3D Image-to-3D API. Quickly generate high-quality 3D models from images using Hitem3D AI services.
README
Hi3D-comfyUI
ComfyUI custom node package for Hi3D Image-to-3D API. Generate high-quality 3D models from images using Hi3D AI services.
Features
- Image to 3D — Generate 3D models (GLB) from a single image or multi-view images.
- Unified scene switch — Use the
sceneparameter to switch between general object models and portrait/scene models in one node. - Model-resolution linkage — The node UI automatically filters available model versions and resolutions based on the selected
sceneandmodel. - Texture regeneration — Re-texture existing 3D models based on a reference image.
- Load 3D model — Load a GLB file from disk with a built-in upload button (GLB only).
Installation
Download from Hi3D Website
- Download the
Hi3D-comfyUIarchive from the Hi3D website or the project releases page. - Extract it into the
ComfyUI/custom_nodesdirectory. - Install Python dependencies via the ComfyUI built-in terminal (click the terminal icon in the ComfyUI Manager toolbar) or any terminal with the same Python environment:
cd ComfyUI/custom_nodes/Hi3D-comfyUI
pip install -r requirements.txt
ComfyUI Manager
Search for Hi3D in ComfyUI Manager and install it directly.
Configuration
An API key pair (ak / sk) is required. Register at the Hi3D website to obtain one.
Option 1: config.json (Recommended)
Edit config.json in the node directory:
{
"Hi3D_ak": "your_access_key",
"Hi3D_sk": "your_secret_key"
}
Option 2: Environment Variables
export Hi3D_ak=your_access_key
export Hi3D_sk=your_secret_key
If ak/sk are configured via
config.jsonor environment variables, the input fields on the nodes can be left empty.
Nodes
Hi3D:ImageTo3D
Generates a 3D model (GLB) from a single image or multi-view images.
| Parameter | Type | Description |
|-----------|------|-------------|
| ak | STRING | API Access Key (leave empty if configured globally) |
| sk | STRING | API Secret Key (leave empty if configured globally) |
| image | IMAGE | Main front image (required) |
| image_back | IMAGE | Back reference image (optional, for multi-view) |
| image_left | IMAGE | Left reference image (optional, for multi-view) |
| image_right | IMAGE | Right reference image (optional, for multi-view) |
| texture | BOOLEAN | Generate texture maps (default: False) |
| pbr | BOOLEAN | PBR 开关(默认开启)。关闭=关闭,开启=开启。PBR需要 texture=True,且仅支持 hitem3dv2.0 / hitem3dv2.1。 |
| scene | COMBO | Scene preset: general / portrait |
| model | COMBO | Model version (filtered by scene, see table below) |
| resolution | COMBO | Generation resolution (filtered by model, default: 1024, i.e. 1024³) |
| face | INT | Target poly count (100,000–2,000,000, supports custom value). UI recommends by resolution: 512→500000, 1024→1000000, 1536/1536fast/1536pro/1536profast→2000000. |
Scene → Model mapping:
| Scene | Available Models |
|-------|-----------------|
| general | hitem3dv1.5, hitem3dv2.0, hitem3dv2.1 |
| portrait | scene-portraitv1.5, scene-portraitv2.0, scene-portraitv2.1 |
Model → Resolution mapping:
| Model | Available Resolutions |
|-------|----------------------|
| hitem3dv1.5 | 512, 1024, 1536, 1536pro |
| hitem3dv2.0 | 512, 1024, 1536, 1536pro |
| hitem3dv2.1 | 1536fast, 1536pro |
| scene-portraitv1.5 | 1536 |
| scene-portraitv2.0 | 1536pro |
| scene-portraitv2.1 | 1536profast, 1536pro |
Outputs:
| Output | Type | Description |
|--------|------|-------------|
| GLB | FILE3DGLB | Downloaded 3D model in GLB format |
| model_task | HI3D_MODEL_TASK | Task info, can be passed to the Texture node |
Hi3D:Texture
Regenerates texture maps for an existing 3D model based on a reference image.
| Parameter | Type | Description |
|-----------|------|-------------|
| ak | STRING | API Access Key (leave empty if configured globally) |
| sk | STRING | API Secret Key (leave empty if configured globally) |
| GLB | FILE3DGLB | Input 3D model in GLB format |
| image | IMAGE | Texture reference image (required) |
| model_task | HI3D_MODEL_TASK | Upstream task info from ImageTo3D (optional) |
| model | COMBO | Model version: hitem3dv1.5 / hitem3dv2.0 / hitem3dv2.1 / scene-portraitv1.5 |
| pbr | BOOLEAN | PBR 开关(默认开启)。关闭=关闭,开启=开启。仅支持 hitem3dv2.0 / hitem3dv2.1。 |
Provide either
model_taskorGLB. Whenmodel_taskis connected, its model URL is used automatically without re-uploading.
Outputs: Same as ImageTo3D.
Hi3D:Load3DModel
Loads a GLB file from disk for use with the Texture node. Includes an Upload GLB File button to browse and upload files from your local computer.
| Parameter | Type | Description |
|-----------|------|-------------|
| model_file | COMBO | Select a GLB file from the dropdown or upload a new one |
Output:
| Output | Type | Description |
|--------|------|-------------|
| model_3d | FILE3D | Loaded 3D model, can be connected to the Texture node's GLB input |
Example Workflows
Image to 3D
Generate a 3D model from a single image:

Texture Regeneration
Re-texture an existing 3D model with a reference image:

License
MIT
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.