Black Forest Lab, founded by the creators of Stable Diffusion, recently launched an exciting text-to-image model — FLUX.1. This new model supports Chinese input and shows excellent language understanding and text control. FLUX.1 offers three image-generation models, two of them open-source, giving users multiple choices.
\ \We compared FLUX.1 Pro with Midjourney's latest v6.1 model. While FLUX.1 lags Midjourney in artistic style and diversity, FLUX.1 Pro matches or even beats v6.1 in image sharpness, detail realism and complex-scene generation. Especially in handling complex English text, multi-person scenes and prompt understanding, FLUX.1 Pro is significantly better. It accurately handles multi-person scenes, avoids facial and hand distortion, renders long text with correct content and layout, and effectively prevents prompt leaking.
\ \

The FLUX.1 all-in-one package is based on Stable Diffusion WebUI forge, using the lightweight nf4 quantized model. Compared with nf8, nf4 achieves better results in a smaller size. Tested on 6GB VRAM, it runs smoothly and takes about 30 seconds per image. After downloading, just unzip and run “启动webui.bat” to open the WebUI, then select FLUX.1's nf4 model to generate.
\ \If your PC can't run FLUX.1, you can still try the text-to-image tool via online platforms. Whether pro, dev or schnell, enter a prompt, set the aspect ratio, and generate the image you want.
\ \\ \
Online experience:
\
https://replicate.com/black-forest-labs