The release of Flux.1 once again set the entire AI mapping field on fire. It is known as the “new benchmark” for AI Vincent maps and has refreshed the new landscape of the AI image field.
Black Forest Lab
The creative team behind Flux is from Black Forest Labs, created by the original team of Stable Diffusion. On August 1, 2024, it was officially announced
The core members include Robin Rombach, who announced his departure from Stability AI in March and is also one of the two main authors of Stable Diffusion
Flux model
Flux has released a total of three image generation models so far, two of which are open source. The three models are
FLUX.1 [pro]
closed source model, which has the best quality and provides state-of-the-art performance in image generation, with top-notch prompt following, visual quality, image details and output diversity. FLUX.1 [pro] can be accessed through APIs and partners Replicate, fal.ai and mystic, supporting enterprise solutions.
FLUX.1 [dev]
FLUX.1 [dev] is an open and lightweight model for non-commercial applications, distilled directly from FLUX.1 [pro], and gains similar quality and timely compliance capabilities as the Pro version, while being more efficient than the standard model of the same size. It can be tried directly on Replicate, fal.ai, mystic and deepinfra.
FLUX.1 [schnell]
Ready for business, tailored for local development and personal use, FLUX.1 [schnell] is publicly available under the Apache 2.0 license. FLUX.1 [schnell] is also available via Replicate, fal.ai, mystic and deepinfra.
Flux ComfyUI usage tutorial
Flux is an open source model that can be deployed locally and on cloud servers using ComfyUI to quickly build workflows and generate images
Configuration requirements: Minimum graphics memory requirement: 8-12GB or higher (rendering speed is slow) Recommended graphics memory: 16-24GB
The requirements for the computer configuration are high, and the model is very large. If your computer is only average, if you are just learning and experiencing Flux, it is not recommended to deploy locally. It is recommended to use a cloud server for quick deployment
Flux online experience
In addition, the official website also provides a series of online experience addresses, which are summarized here
Description: Free trial available, new users get $1 free, can try dev, can generate up to 40 images
Hugging Face flux dev
Price: Free, available with FLUX.1 [Dev] and FLUX.1 [schnell
Price: Free trial available, available with FLUX.1 [Pro], FLUX.1 [Dev] and FLUX.1 [schnell
Description: Paid use, no free trial, not recommended. flux-pro: $0.055 per image flux-dev: $0.03 per image
flux-schnell: $0.003 per image
Flux vs Midjourney
Text generation results
Prompt1: detailed cinematic dof render of an old dusty detailed CRT monitor on a wooden desk in a dim room with items around, messy dirty room. On the screen are the letters “FLUX” glowing softly. High detail hard surface render.
Prompt2: Photograph of a blackboard in an old classroom. On the blackboard in chalk are the words “Let’s make some really pretty stuff together” with a red chalk heart after the words. Sunlight is streaming in from the window
Prompt3: Portrait of an old man with glasses, 50mm, f1.4, natural light, Pathéchrome
Prompt4: Close-Up black and white Shot of a woman looks at the camera by Olivier Theyskens
Anime style effect
Prompt5: beautiful anime artwork, a cute anime girl standing alone at night darkness wearing an oversized black raincoat with a small bag. she is holding a black umbrella, the umbrella has multicolored LED hidden inside that shine down on her, small rose petals flutter through the air around her, anime screencap style, red eyes, medium hair.
Landscape generation effect
Prompt 5: A photo of the beach in Green Cay, lit by pastel sunset colors, bright. The sky is dark and cloudy with some pink hues on top. The water reflects these hues, creating a beautiful scene. The sea looks calm and peaceful, and the horizon line has just shrunk to show that we’re looking at an early morning sunrise. Ultra-detailed, intricate details, HD, best quality, high contrast, best quality, high detail, 16k.
Overall, some of the performance, including the effectiveness of following prompt words, the accuracy of text generation effects, and size control, has surpassed Midjourney to some extent, but there is still some distance to go in terms of the diversity of the generated style effects and the overall aesthetic effect of the screen.
The above is a brief comparison of some of the best versions of the two models in some directions. The generated effects are for your reference
Summary
The release of Flux.1 has re-invigorated the AI drawing field, which had been quiet for half a year, and Flux has also ascended to the throne of the strongest open source Vincent model.
However, BlackForestLabs’ ambitions do not stop with the Vincent model, and the next stage is to enter the field of Vincent video!