README
Note for Hunyuan3D-2
Looking for pre-built portables of Hunyuan3D 2.0 ? Check out:
What Can It Do?
This build, pt25-cu124, is based on PyTorch 2.5.1 with CUDA 12.4,
targeting the
Dec 18, 2024
version of ComfyUI-3D-Pack.
I tested all the example workflows on a fairly-old TITAN Xp GPU (12GB VRAM). The results are as follows:
| Workflow | Status | Screenshot | Memo |
|---|---|---|---|
(NEW) Trellis_Image_to_Mesh | ๐ | Default to xFormers for compatibility. To enable Flash Attention, see Make Your Own. | |
(NEW) Hunyuan3D_V1 | ๐ | Default to xFormers for compatibility. To enable Flash Attention, see Make Your Own. | |
(NEW) StableFast3D / sf3d_by_stability_ai | ๐ | Workflow from Stability AI. | |
StableFast3D_to_Mesh | ๐ | Workflow from 3D-Pack. | |
CharacterGen_Vanilla | ๐ | For compatibility with CharacterGen Loader, an older version of | |
CharacterGen_to_Unique3D | ๐ | Workflow executes successfully, however the generated mesh is broken. | |
CharacterGen_InstantMesh | โ | CharacterGen: ๐; InstantMesh: OOM. | |
CRM MVDiffusion_to_Craftsman | ๐ | For compatibility with CRM Loader, an older version of | |
CRM MVDiffusion_to_Unique3D | ๐ | ||
CRM_Img_to_Multi-views_to_CCMs_to_Mesh | ๐ | ||
CRM MVDiffusion_to_InstantMesh | โ | Incompatible InstantMesh node. | |
CRM_T2I_V2 | โ | ||
Unique3D_All_Stages | ๐ | As well as Unique3D stage 0 to 3. | |
Unique3D_with_Era3D_All | ๐ | Previously incompatible, now working. | |
[FlexiCubes] Multi-View Depth plus Optional Normal maps to 3D Mesh | ๐ | ||
Era3D_to_InstantMesh | โ | Era3D: ๐; InstantMesh: OOM or cube. | |
Multi-View-Images_to_Instant-NGP_to_ DiffRast_to_3DMesh | ๐ | Previously incompatible, now working. | |
MVDream_to_Large_Multiview_Gaussian_ Model_to_3DGS and_3DMesh | ๐ | Previously incompatible, now working. | |
Render_Mesh_and_3DGS_Example | ๐ | ||
Triplane_Gaussian_Transformers_to_ 3DGS_and_3DMesh | ๐ | Previously incompatible, now working. | |
TripoSR_to_Mesh | ๐ | The "easy" model that always works. | |
Using_Stack_Orbit_Camera_Poses_ to_conditioning_StableZero123 | ๐ | ||
Wonder3D_to_Large_Multiview_ Gaussian_Model_to_3DGS and_3DMesh | ๐ | Previously incompatible, now working. | |
Zero123Plus_to_InstantMesh | ๐ |
-
Compared to the previous
pt24-cu121build, the newpt25-cu124shows no regression in compatibility. All workflows that worked previously continue to function in the new build.
Before You Startโฆโ
-
Know your GPUโs CUDA compute capability (arch number). For example, RTX 2080 is
7.5, RTX 3090 is8.6and RTX 4090 is8.9. You can find the number in NVIDIA doc or this article. -
(Optional) If you want to use Stable Fast 3D, you need a Hugging Face access token to download it.
-
Register/Login Hugging Face.
-
Go to Stable Fast 3D repository, fill in and agree Stability AIโs License Agreement.
-
Go to Access Tokens, create a read-only token. Save it somewhere for later use.
-
Download, Install, Compile and Run
1. Basic Requirements
1.1. Install CUDA Toolkit 12.4.1.
-
Only Libraries and Compilers are needed to install.
-
If your GPU driver is outdated (<550), you can install the driver provided by the CUDA Toolkit installer. Otherwise, you donโt need to downgrade the driver version.
-
During the setup, select "Desktop development with C++" workload.
-
(Just in case) If new major version of Visual Studio is released, you need to manually select version 17.x (e.g. 17.12) and MSVC v143. Referring to CUDA 12.4 compiler support.
2. Download and Config
2.1. Download the package files from the release page:
-
https://github.com/YanWenKun/Comfy3D-WinPortable/releases/tag/r7-pt25
-
Note that there are multiple files (compressed in volumes) that need to be downloaded. Extract it to a convenient place.
2.2. Edit !FIRST_RUN-compile-install.bat
You can see the line:
set TORCH_CUDA_ARCH_LIST=6.1+PTX
Edit it according to your GPU arch.
For example, for an RTX 4060 Ti, it should be:
set TORCH_CUDA_ARCH_LIST=8.9
2.3. Edit run.bat
You can see the line (again):
set TORCH_CUDA_ARCH_LIST=6.1+PTX
Edit it like you just did in the last step.
3. (Optional) Extra Config
3.1. There are few more settings available in the start script run.bat, such as:
-
Set proxy
-
Do not open browser after starting
Each line of configuration has a description above it.
If needed, uncomment the according line (remove rem at the beginning),
edit it and save the file.
3.2. If you are going to run Stable Fast 3D workflows, you need to edit the start script run.bat, find the line:
rem set HF_TOKEN=
and edit it to:
set HF_TOKEN=hf_your_token
For example:
set HF_TOKEN=hf_ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefgh
3.3. If you run multiple ComfyUI instances in Sandboxie, as a reminder, the .bat scripts should be executed in a sandbox.
In addition, I recommend to set the program folder (Comfy3D_WinPortable) to "Full Access" under "Sandbox Settings - Resource Access", to improve disk I/O.
4. First Start
4.1. Run !FIRST_RUN-compile-install.bat. Wait until it finish installation.
-
Depending on hardware, it may take 5~30 minutes.
-
If it fails, simply run the script again.
4.2. If you are going to run Unique3D workflows, run download-models-for-Unique3D.bat.
4.3. Run run.bat. After starting, the program will automatically open a browser, or you can visit: http://localhost:8188/
-
In the left sidebar, you can find the "Workflows" button, the example workflows are listed there. The example input files are already in the
inputfolder. Open a workflow and click "Queue" at bottom of the page to run the workflow. -
To shutdown the program, simply close the console window.
-
To start the program in the future, simply run
run.bat.
Troubleshooting
A. Force-Update
It is not recommended to update any component of this package, whether itโs Python, ComfyUI or Custom Nodes ("If it works, donโt fix it").
But if you really need to update ComfyUI and Custom Nodes (except 3D-Pack), thereโs a script to do that. Run this in your Git Bash:
bash force-update-all.sh
This script will force git-pull all the repositories.
Note that ComfyUI\custom_nodes\ComfyUI-3D-Pack folder is not a git repository, so it wonโt be affected by this script.
B. Re-Compile and Re-Install Dependencies
Just like the FIRST_RUN script did,
the reinstall-deps-for-3d.bat script will attempt to compile and install the dependencies of 3D-Pack.
You need to edit this file and set TORCH_CUDA_ARCH_LIST as well.
There are few differences from the FIRST_RUN, this script will:
-
Download the latest source repos, not the targeted versions. This may help with compatibility issues when the FIRST_RUN doesnโt work.
-
Additionally compile-install
kiuikit,utils3dandnvdiffrast. In my tests, itโs not necessary to do so in the FIRST_RUN. I put them here for fail-safe. -
Keep temporary files and save artifacts (.whl files) in the working folder
tmp_build, allowing you to reuse them. Keep in mind that they are not essentially portable.
C. Last and Guaranteed - Linux
If you run into complicated compatibility issues, try Linux.
Everything in Comfy3D works more smoothly on Linux. Most 3D workflows are research-oriented and far from production-ready. By installing & running them youโre essentially replicating the development environment used by researchers. Many components need to be compiled before running, and this is exactly where Linux excels.
Check out Dockerfile or my Docker image if youโre familiar with containers.
Make Your Own
You can build this package using GitHub Actions. Simply fork this repo and go to "Actions" page, find "Build & Upload pt25-cu124 Package" and then find "Run workflow".
The GitHub workflow takes about 20 minutes to finish. After that, you can find the 7z files in the "Releases" page.
Examples of custom modification:
-
If you never use CharacterGen, you can delete
diffusers==0.29.1inpakB.txt. -
If you never use CRM, you can delete
open-clip-torch==2.24.0inpakB.txt. -
If you are using Ampere or later GPUs, you can enable Flash Attention in
pak8.txt.
Contributors
YanWenKun
163 commits
Similar repos
Languages
Python
61.9%
Batchfile
22.5%
Shell
15.6%
README
Note for Hunyuan3D-2
Looking for pre-built portables of Hunyuan3D 2.0 ? Check out:
What Can It Do?
This build, pt25-cu124, is based on PyTorch 2.5.1 with CUDA 12.4,
targeting the
Dec 18, 2024
version of ComfyUI-3D-Pack.
I tested all the example workflows on a fairly-old TITAN Xp GPU (12GB VRAM). The results are as follows:
| Workflow | Status | Screenshot | Memo |
|---|---|---|---|
(NEW) Trellis_Image_to_Mesh | ๐ | Default to xFormers for compatibility. To enable Flash Attention, see Make Your Own. | |
(NEW) Hunyuan3D_V1 | ๐ | Default to xFormers for compatibility. To enable Flash Attention, see Make Your Own. | |
(NEW) StableFast3D / sf3d_by_stability_ai | ๐ | Workflow from Stability AI. | |
StableFast3D_to_Mesh | ๐ | Workflow from 3D-Pack. | |
CharacterGen_Vanilla | ๐ | For compatibility with CharacterGen Loader, an older version of | |
CharacterGen_to_Unique3D | ๐ | Workflow executes successfully, however the generated mesh is broken. | |
CharacterGen_InstantMesh | โ | CharacterGen: ๐; InstantMesh: OOM. | |
CRM MVDiffusion_to_Craftsman | ๐ | For compatibility with CRM Loader, an older version of | |
CRM MVDiffusion_to_Unique3D | ๐ | ||
CRM_Img_to_Multi-views_to_CCMs_to_Mesh | ๐ | ||
CRM MVDiffusion_to_InstantMesh | โ | Incompatible InstantMesh node. | |
CRM_T2I_V2 | โ | ||
Unique3D_All_Stages | ๐ | As well as Unique3D stage 0 to 3. | |
Unique3D_with_Era3D_All | ๐ | Previously incompatible, now working. | |
[FlexiCubes] Multi-View Depth plus Optional Normal maps to 3D Mesh | ๐ | ||
Era3D_to_InstantMesh | โ | Era3D: ๐; InstantMesh: OOM or cube. | |
Multi-View-Images_to_Instant-NGP_to_ DiffRast_to_3DMesh | ๐ | Previously incompatible, now working. | |
MVDream_to_Large_Multiview_Gaussian_ Model_to_3DGS and_3DMesh | ๐ | Previously incompatible, now working. | |
Render_Mesh_and_3DGS_Example | ๐ | ||
Triplane_Gaussian_Transformers_to_ 3DGS_and_3DMesh | ๐ | Previously incompatible, now working. | |
TripoSR_to_Mesh | ๐ | The "easy" model that always works. | |
Using_Stack_Orbit_Camera_Poses_ to_conditioning_StableZero123 | ๐ | ||
Wonder3D_to_Large_Multiview_ Gaussian_Model_to_3DGS and_3DMesh | ๐ | Previously incompatible, now working. | |
Zero123Plus_to_InstantMesh | ๐ |
-
Compared to the previous
pt24-cu121build, the newpt25-cu124shows no regression in compatibility. All workflows that worked previously continue to function in the new build.
Before You Startโฆโ
-
Know your GPUโs CUDA compute capability (arch number). For example, RTX 2080 is
7.5, RTX 3090 is8.6and RTX 4090 is8.9. You can find the number in NVIDIA doc or this article. -
(Optional) If you want to use Stable Fast 3D, you need a Hugging Face access token to download it.
-
Register/Login Hugging Face.
-
Go to Stable Fast 3D repository, fill in and agree Stability AIโs License Agreement.
-
Go to Access Tokens, create a read-only token. Save it somewhere for later use.
-
Download, Install, Compile and Run
1. Basic Requirements
1.1. Install CUDA Toolkit 12.4.1.
-
Only Libraries and Compilers are needed to install.
-
If your GPU driver is outdated (<550), you can install the driver provided by the CUDA Toolkit installer. Otherwise, you donโt need to downgrade the driver version.
-
During the setup, select "Desktop development with C++" workload.
-
(Just in case) If new major version of Visual Studio is released, you need to manually select version 17.x (e.g. 17.12) and MSVC v143. Referring to CUDA 12.4 compiler support.
2. Download and Config
2.1. Download the package files from the release page:
-
https://github.com/YanWenKun/Comfy3D-WinPortable/releases/tag/r7-pt25
-
Note that there are multiple files (compressed in volumes) that need to be downloaded. Extract it to a convenient place.
2.2. Edit !FIRST_RUN-compile-install.bat
You can see the line:
set TORCH_CUDA_ARCH_LIST=6.1+PTX
Edit it according to your GPU arch.
For example, for an RTX 4060 Ti, it should be:
set TORCH_CUDA_ARCH_LIST=8.9
2.3. Edit run.bat
You can see the line (again):
set TORCH_CUDA_ARCH_LIST=6.1+PTX
Edit it like you just did in the last step.
3. (Optional) Extra Config
3.1. There are few more settings available in the start script run.bat, such as:
-
Set proxy
-
Do not open browser after starting
Each line of configuration has a description above it.
If needed, uncomment the according line (remove rem at the beginning),
edit it and save the file.
3.2. If you are going to run Stable Fast 3D workflows, you need to edit the start script run.bat, find the line:
rem set HF_TOKEN=
and edit it to:
set HF_TOKEN=hf_your_token
For example:
set HF_TOKEN=hf_ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefgh
3.3. If you run multiple ComfyUI instances in Sandboxie, as a reminder, the .bat scripts should be executed in a sandbox.
In addition, I recommend to set the program folder (Comfy3D_WinPortable) to "Full Access" under "Sandbox Settings - Resource Access", to improve disk I/O.
4. First Start
4.1. Run !FIRST_RUN-compile-install.bat. Wait until it finish installation.
-
Depending on hardware, it may take 5~30 minutes.
-
If it fails, simply run the script again.
4.2. If you are going to run Unique3D workflows, run download-models-for-Unique3D.bat.
4.3. Run run.bat. After starting, the program will automatically open a browser, or you can visit: http://localhost:8188/
-
In the left sidebar, you can find the "Workflows" button, the example workflows are listed there. The example input files are already in the
inputfolder. Open a workflow and click "Queue" at bottom of the page to run the workflow. -
To shutdown the program, simply close the console window.
-
To start the program in the future, simply run
run.bat.
Troubleshooting
A. Force-Update
It is not recommended to update any component of this package, whether itโs Python, ComfyUI or Custom Nodes ("If it works, donโt fix it").
But if you really need to update ComfyUI and Custom Nodes (except 3D-Pack), thereโs a script to do that. Run this in your Git Bash:
bash force-update-all.sh
This script will force git-pull all the repositories.
Note that ComfyUI\custom_nodes\ComfyUI-3D-Pack folder is not a git repository, so it wonโt be affected by this script.
B. Re-Compile and Re-Install Dependencies
Just like the FIRST_RUN script did,
the reinstall-deps-for-3d.bat script will attempt to compile and install the dependencies of 3D-Pack.
You need to edit this file and set TORCH_CUDA_ARCH_LIST as well.
There are few differences from the FIRST_RUN, this script will:
-
Download the latest source repos, not the targeted versions. This may help with compatibility issues when the FIRST_RUN doesnโt work.
-
Additionally compile-install
kiuikit,utils3dandnvdiffrast. In my tests, itโs not necessary to do so in the FIRST_RUN. I put them here for fail-safe. -
Keep temporary files and save artifacts (.whl files) in the working folder
tmp_build, allowing you to reuse them. Keep in mind that they are not essentially portable.
C. Last and Guaranteed - Linux
If you run into complicated compatibility issues, try Linux.
Everything in Comfy3D works more smoothly on Linux. Most 3D workflows are research-oriented and far from production-ready. By installing & running them youโre essentially replicating the development environment used by researchers. Many components need to be compiled before running, and this is exactly where Linux excels.
Check out Dockerfile or my Docker image if youโre familiar with containers.
Make Your Own
You can build this package using GitHub Actions. Simply fork this repo and go to "Actions" page, find "Build & Upload pt25-cu124 Package" and then find "Run workflow".
The GitHub workflow takes about 20 minutes to finish. After that, you can find the 7z files in the "Releases" page.
Examples of custom modification:
-
If you never use CharacterGen, you can delete
diffusers==0.29.1inpakB.txt. -
If you never use CRM, you can delete
open-clip-torch==2.24.0inpakB.txt. -
If you are using Ampere or later GPUs, you can enable Flash Attention in
pak8.txt.
Similar repos
Contributors
YanWenKun
163 commits
Languages
Python
61.9%
Batchfile
22.5%
Shell
15.6%