Frank Schwenk
35 lines
1.2 KiB
Markdown
35 lines
1.2 KiB
Markdown
# ComfyUI Flux Klein img2img workflow (API format)
|
|
|
|
Export a Flux Klein img2img workflow from ComfyUI once and save it as:
|
|
|
|
```
|
|
workflows/comfy/flux_klein_edit_api.json
|
|
```
|
|
|
|
Copy `flux_klein_edit_api.json.example` as a starting skeleton, then replace node IDs and wiring with your exported workflow.
|
|
|
|
## Export steps
|
|
|
|
1. Build an img2img workflow in ComfyUI with **Load Image**, prompt encoding, sampler, and **Save Image** nodes.
|
|
2. Enable **Dev mode** in ComfyUI settings if needed.
|
|
3. Use **Save (API Format)** on the workflow.
|
|
4. Save the JSON to `workflows/comfy/flux_klein_edit_api.json`.
|
|
|
|
## Nodes patched by `comfy_flux_edit`
|
|
|
|
The module walks workflow nodes by `class_type` and updates:
|
|
|
|
| class_type | Field | Value |
|
|
|------------|-------|-------|
|
|
| `LoadImage` | `inputs.image` | Uploaded filename |
|
|
| `CLIPTextEncode` / `TextEncode` | `inputs.text` | Step `prompt` |
|
|
| `KSampler` | `inputs.denoise`, `inputs.seed` | Step params |
|
|
|
|
## CPU warning
|
|
|
|
ComfyUI on CPU can take hours per full-resolution image. Use `max_edge=1024` or lower in the pipeline step for experiments.
|
|
|
|
## Server
|
|
|
|
Default URL: `http://127.0.0.1:8188`. Start ComfyUI before running the pipeline step.
|