FAQ

Can I use ComfyUI which is included in Visionatrix?

Yes, you can install your Nodes there and run ComfyUI separately without running Visionatrix.

Also, ComfyUI is available from the Visionatrix UI or at http://127.0.0.1:8188.

Can I use my own already installed ComfyUI with Visionatrix?

By default, Visionatrix installs or expects ComfyUI to be in working directory under the ComfyUI folder name.

You can also tell Visionatrix to use ComfyUI installed in a different path, by setting the COMFYUI_DIR environment variable.

Also, you can set comfyui_folder database parameter(by using set-global-setting command) if you do not want to set the environment variable.

---

Can I run it on multiple GPU?

You can run one worker on one GPU and process tasks in parallel, take a look at Server and Worker modes.

---

Why are portrait flows on runpod.io / novita.ai slow?

Portrait flows on platforms like runpod.io or novita.ai often run slowly because these setups typically use Docker containers with restricted CPU instructions, causing the onnxruntime library to default to the slower CPUExecutionProvider.

To significantly improve performance, you can set the execution provider explicitly to use GPU acceleration with CUDA. Execute the following command:

python3 -m visionatrix set-global-setting --key="insightface_provider" --value="CUDA"

This overrides the default provider, allowing the flows to leverage GPU hardware and run faster.

---