Getting Started with ComfyUI on Thumper-Run

Install ComfyUI the Easy Way

Most guides tell you to clone a repo, install Python, manage virtual environments, and pray your CUDA version matches. With Thumper-Run, you skip all of that. Open the app catalog, find ComfyUI, and click Install. Thumper handles Python isolation, dependency pinning, and GPU detection automatically. The entire process takes two to five minutes depending on your internet connection.

If you hit a snag, the built-in health monitor will surface the exact error and suggest a fix. No more hunting through terminal output.

Load Your First Workflow

ComfyUI ships with a default workflow that is perfect for a first test. When you launch ComfyUI through Thumper, the default graph appears: a KSampler node connected to a CLIP Text Encode prompt, a VAE Decode node, and a Save Image output. You do not need to build anything from scratch.

Want something more advanced? Thumper bundles curated starter workflows you can load from the Workflow Library tab. Each one is tested against the model packs available in the catalog, so you will never see a missing-node error on first run.

Generate Your First Image

Type a prompt into the positive CLIP node—something like *"a cozy cabin in a snowy forest, cinematic lighting"*—and click Queue Prompt. ComfyUI processes the graph node-by-node. On a mid-range GPU (RTX 3060, RX 7600) you should see a 512×512 image in under 15 seconds.

The image is saved to your local output directory and stays local unless you explicitly configure a sharing, sync, or cloud workflow. The security overview explains the boundary for each network feature.

Understanding Nodes

ComfyUI is a node-based workflow editor. Each node performs one operation:

• Load Checkpoint – loads a Stable Diffusion model from disk
• CLIP Text Encode – converts your text prompt into embeddings
• KSampler – runs the diffusion sampling loop
• VAE Decode – converts the latent image into pixels
• Save Image – writes the result to disk

Nodes connect via colored wires. Purple carries model data, yellow carries latents, and pink carries images. Once you internalize this, you can build any pipeline—inpainting, ControlNet, LoRA stacking—by snapping nodes together.

Model Packs

Thumper provides Model Packs so you do not have to hunt for compatible checkpoints. A pack bundles the base model, VAE, and any required LoRA weights into a single download. Install an SDXL pack and every bundled workflow will resolve its models automatically.

See the ComfyUI guide for a list of recommended packs and their VRAM requirements.

Next Steps

• Browse the workflow library for advanced pipelines
• Read the GPU setup guide if you are on AMD hardware
• Explore CRDT sync to keep workflows in sync across devices

Ready to try it? Download Thumper-Run free →