crispz-klein is a Fooocus-style web studio for FLUX.2 Klein 4B, Black Forest Labs' Apache 2.0 model distilled to 4 steps. One model covers the whole surface: text to image, multi-reference editing and inpaint run in the same pipeline, with no second model loaded. Measured on an RTX 5090: 14.9 GB of VRAM for everything, 1024x1024 in 2.0 s, an edit with one reference in 3 s.
| Generate |
Edit (1 reference) |
Inpaint |
 |
 |
 |
The middle image is the point: "make it daytime, sunny blue sky", and the character stays pixel-identical, same face, same pose, same clothes.
- Edit with up to 4 references in the tab (more through the CLI), two verified edit-LoRA presets (Consistence-Edit for the 4B and the 9B).
- Inpaint, outpaint, img2img in the same editor, ESRGAN + refine upscale, hand and face detailer.
- The family toolbox: styles, wildcards, tag autocomplete, LoRAs with hot swap, CivitAI search, job queue, X/Y/Z grids, Describe / Improve / Vision Mix through Ollama, Remove BG, Face Swap, CLI and the protocol v1 (
czp).
- Honest about the model: klein is step-wise distilled, so negative prompts and the guidance slider have no effect (verified bit-identical renders at 1, 4 and 8); the UI says so instead of pretending.
Hardware: an NVIDIA GPU with 16 GB of VRAM is comfortable. The weights are public on Hugging Face: no licence to accept, no token.
Pinokio launcher: https://github.com/mikecastrodemaria/crispz-klein.pinokio
Source: https://github.com/mikecastrodemaria/crispz-klein
Install (Pinokio)
- In Pinokio, open Discover and paste
https://github.com/mikecastrodemaria/crispz-klein.pinokio.
- Click Install: the launcher clones the app, creates the venv, installs the requirements and PyTorch (CUDA 12.8 on NVIDIA).
- Click Start, then Open Web UI. The first render downloads the model.
crispz-klein is a Fooocus-style web studio for FLUX.2 Klein 4B, Black Forest Labs' Apache 2.0 model distilled to 4 steps. One model covers the whole surface: text to image, multi-reference editing and inpaint run in the same pipeline, with no second model loaded. Measured on an RTX 5090: 14.9 GB of VRAM for everything, 1024x1024 in 2.0 s, an edit with one reference in 3 s.
The middle image is the point: "make it daytime, sunny blue sky", and the character stays pixel-identical, same face, same pose, same clothes.
czp).Hardware: an NVIDIA GPU with 16 GB of VRAM is comfortable. The weights are public on Hugging Face: no licence to accept, no token.
Pinokio launcher: https://github.com/mikecastrodemaria/crispz-klein.pinokio
Source: https://github.com/mikecastrodemaria/crispz-klein
Install (Pinokio)
https://github.com/mikecastrodemaria/crispz-klein.pinokio.