ComfyUI Node
Flux Klein img2img ⚡
Prompts, init latent, and one reference image for FLUX.2 Klein. Leave image unconnected for txt2img. Feed the outputs straight into a stock KSampler.
Flux Klein img2img ⚡
- model
- clip
- vae
- image
- reference_image
- model
- positive
- negative
- latent
- denoise
◄prompt—►
◄negative_prompt►
◄strength0.60►
◄batch_size1►
◄width1024►
◄height1024►
◄control_modemanual►
◄depth_ckpt_namedepth_anything_v2_vitb.pth►
Category🤖 CCTech/Flux Klein
Inputs (13)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | — | |
| clip | CLIP | — | |
| vae | VAE | — | |
| prompt | STRING | — | |
| negative_prompt | STRING | — | |
| strength | FLOAT | 0.600–1 | img2img only. How much of the init image to discard. Ignored without an image. |
| batch_size | INT | 11–4096 | — |
| width | INT | 102416–16384 | — |
| height | INT | 102416–16384 | — |
| imageopt | IMAGE | Init image for img2img. Leave unconnected for txt2img. | |
| reference_imageopt | IMAGE | Klein's real edit mechanism: encoded and attached to positive+negative conditioning as reference_latents. Independent of `image` - for a pure reference-driven edit, leave `image` unconnected (txt2img latent) and connect only this. | |
| control_modeopt | COMBO | manual | manual: attach reference_image raw. auto_depth: run it through Depth Anything V2 first and attach the depth map instead - reproduces the real example workflow's structural-reference trick (AIO_Preprocessor -> MiDaS depth -> reference_latents). |
| depth_ckpt_nameopt | COMBO | depth_anything_v2_vitb.pth | auto_depth mode only. Model size for the automatic depth estimation. Downloads on first use if not already in models/depth_anything_v2. |
Outputs (5)
| Name | Type | Description |
|---|---|---|
| model | MODEL | — |
| positive | CONDITIONING | — |
| negative | CONDITIONING | — |
| latent | LATENT | — |
| denoise | FLOAT | — |