ComfyUI Setup Guide

This guide is for users who already have ComfyUI installed locally.

Recommended Path

1. Start Museboard.
2. Open 设置.
3. Find ComfyUI 连接向导.
4. Leave the endpoint as:

http://127.0.0.1:8188

1. Click 测试连接.
2. If connection fails, enable auto start and set:
   • ComfyUI working directory
   • ComfyUI launch command
3. Click 启动并连接.
4. Click Flux 推荐预设.
5. Click 生成测试图.

If the test image appears, ComfyUI is ready.

Auto Start Settings

Default working directory:

~/ComfyUI

Default launch command:

.venv/bin/python main.py --listen 127.0.0.1 --port 8188

Museboard intentionally blocks shell syntax such as ;, |, &, $, and redirection. The launch command must be a program plus arguments only.

Flux Preset

The Flux preset asks ComfyUI for available node and model information, then builds a workflow automatically.

It expects ComfyUI to expose compatible node types such as:

• UNETLoader
• CLIPLoader
• VAELoader
• Flux2Scheduler
• EmptyFlux2LatentImage
• CFGGuider
• ConditioningZeroOut
• SaveImage

If one of these is missing, Museboard will report which part failed. The fix is usually installing the matching ComfyUI custom nodes or using a workflow imported from your own ComfyUI setup.

Importing Your Own Workflow

Use ComfyUI’s API workflow export, not the normal UI workflow export.

After importing JSON, Museboard tries to identify:

• positive prompt node
• negative prompt node
• width node
• height node
• seed node
• SaveImage node

If auto-mapping is wrong, open 高级 workflow 与节点映射 and correct the fields manually.

Common Failures

Cannot connect

Likely causes:

• ComfyUI is not running.
• The endpoint is not http://127.0.0.1:8188.
• ComfyUI launched on a different port.
• Browser preview cannot auto-start local processes. Use the desktop app.

Flux preset fails

Likely causes:

• Flux model files are missing.
• Required custom nodes are missing.
• The installed ComfyUI version exposes different node names.

Use 复制诊断信息 in the ComfyUI setup wizard and inspect the node classes listed in the report.

Workflow JSON fails

Likely causes:

• The file is not API workflow JSON.
• Node IDs changed after editing the workflow.
• The positive prompt, width, or height nodes were not mapped.

Open 高级 workflow 与节点映射 and verify node IDs.

Generation starts but fails

Likely causes:

• VRAM/RAM is insufficient.
• The selected image size is too large.
• A model referenced by the workflow is missing.
• A custom node raised an error.

Reduce image size first. If that fails, copy diagnostics and check the ComfyUI terminal log.