license: apache-2.0
pipeline_tag: image-to-image
tags:
- flux
- inpainting
- image-editing
- diffusion
- safetensors
base_model: black-forest-labs/FLUX.2-klein-4B
FLUX Klein 4B — AEmotionStudio Mirror
Mirror of black-forest-labs/FLUX.2-klein-4B for use with ComfyUI-FFMPEGA.
About
FLUX Klein is a 4-billion parameter image editing and generation model from Black Forest Labs. It supports:
- Object removal — erase unwanted objects from images/video frames
- Text-guided editing — change hair color, swap backgrounds, modify clothing, etc.
- Inpainting — fill masked regions with contextually appropriate content
This mirror provides the same bf16 safetensors weights as the upstream repo, hosted on AEmotionStudio infrastructure for supply-chain resilience.
Model Details
| Property | Value |
|---|---|
| Architecture | FLUX (Rectified Flow Transformer) |
| Parameters | 4B |
| Precision | bf16 (bfloat16) |
| Total Size | ~15 GB |
| License | Apache 2.0 |
| Source | black-forest-labs/FLUX.2-klein-4B |
Files
| Component | File | Size |
|---|---|---|
| Transformer | transformer/diffusion_pytorch_model.safetensors |
7.4 GB |
| Text Encoder (T5-XXL) | text_encoder/model-00001-of-00002.safetensors |
4.7 GB |
| Text Encoder (T5-XXL) | text_encoder/model-00002-of-00002.safetensors |
2.9 GB |
| VAE | vae/diffusion_pytorch_model.safetensors |
160 MB |
| Tokenizer, configs | tokenizer/, scheduler/, *.json |
< 15 MB |
Usage with ComfyUI-FFMPEGA
This model is auto-downloaded when you use the auto_mask skill with effect=remove or effect=edit in ComfyUI-FFMPEGA.
Object Removal
Prompt the FFMPEGA Agent:
Remove the watermark
Or use the Effects Builder with SAM3 target + remove effect.
Text-Guided Editing
Prompt the FFMPEGA Agent:
Change the hair to red
Or use auto_mask with effect=edit and edit_prompt="change the hair to red".
Manual Download
If auto-download is disabled, download all files and place them in:
ComfyUI/models/flux_klein/
VRAM Requirements
- Minimum: ~8 GB (with aggressive CPU offloading)
- Recommended: ~13 GB (bf16/fp16, sequential CPU offload)
- GPUs without native bf16 support (pre-Ampere) automatically use fp16 at load time
License
The model weights are licensed under Apache 2.0 by Black Forest Labs. See LICENSE.md for full terms.
Citation
@misc{flux2klein2025,
title={FLUX.2 Klein},
author={Black Forest Labs},
year={2025},
url={https://huggingface.co/black-forest-labs/FLUX.2-klein-4B}
}