Before you start
Deploy an available ComfyUI template and allocate enough persistent storage for models, inputs, and outputs. Third-party models have their own licenses.
Steps
- Wait for application readiness and open ComfyUI from the GPU instance Connections section.
- If HTTP Basic authentication is requested, use username gpu and the rental’s application token on the HTTPS endpoint. Your console API key is not the password.
- Open a workflow template that matches the installed nodes and selected GPU memory.
- Check Workflow Overview for missing models and nodes before pressing Run. Install required models into the container using the download controls supported by the certified image.
- Begin with a small workflow and conservative resolution or duration. Run it, inspect the queue, and save the generated output.
Check the result
The workflow completes and output appears in the application and persistent workspace.
Troubleshooting
If authentication repeats, reopen the current Connections link and confirm you are using current rental credentials. If a workflow fails, check missing files, unsupported nodes, disk space, and GPU memory before restarting.