Quick Start¶
Deploy ComfyUI Studio on RunPod and open the web UI in under 5 minutes.
Step 1: Deploy on RunPod¶
Click the deploy button or go to RunPod Deploy.
Select a GPU. Any NVIDIA GPU from V100 to B200 works. Recommended:
| Use Case | Recommended GPU | VRAM |
|---|---|---|
| Video generation (WAN 2.2, HunyuanVideo) | A100 80GB, H100, B200 | 80+ GB |
| Image generation (Flux, SDXL) | RTX 4090, A6000, L40S | 24-48 GB |
| Budget / experimentation | RTX 3090, A5000 | 24 GB |
Step 2: Set Environment Variables¶
Before launching the pod, set these environment variables in the RunPod template:
| Variable | Required | Value |
|---|---|---|
API_KEY |
Yes | Choose a strong password. This is your login to the web UI. |
CIVITAI_API_KEY |
Recommended | Your CivitAI API token. Without this, you cannot download models from CivitAI or fetch model metadata. Get it from civitai.com/user/account → API Keys. |
HF_TOKEN |
Recommended | Your HuggingFace access token. Without this, you cannot download gated models (Flux, WAN, Hunyuan, etc.). Get it from huggingface.co/settings/tokens → New token → Read access. |
API_KEY is required
The default API_KEY is changeme. You must set your own strong password. Anyone with this password has full access to your pod through the web UI.
Step 3: Launch and Wait¶
Launch the pod. The first boot takes 2-5 minutes because:
- The bootstrap script clones the application from GitHub
- ComfyUI is copied from the Docker image to the persistent volume
- Both services start (ComfyUI on 8188, Studio on 8000)
You can watch the boot progress in the RunPod container logs. Look for:
=== ComfyUI Studio — starting ===
[0/3] First boot — bootstrapping application...
[bootstrap] Cloning repo...
[bootstrap] Copied backend
[bootstrap] Copied frontend
[bootstrap] Bootstrap complete — app v12.3.0
[1/3] ComfyUI already present — skipping
[2/3] Starting ComfyUI on port 8188...
ComfyUI ready after 10 seconds.
[3/3] Starting Studio backend on port 8000...
=== Services started ===
Step 4: Open the Web UI¶
Once the services are running, open:
Replace <POD_ID> with your pod's ID (shown in the RunPod dashboard).
You'll see a login page. Enter the API_KEY you set in step 2.
Step 5: You're In¶
After login, you'll see the dashboard with:
- Component versions and update status
- GPU/CPU/RAM/disk telemetry
- Quick stats (models, workflows, nodes)
From here:
- Go to Models to download AI models
- Go to Workflows to see available generation pipelines
- Go to Run to execute a workflow
Next Steps¶
- Your First Generation — download a model and generate something
- Environment Variables — all configuration options
- Architecture Overview — understand how the system works