The free FLUX model generates images that are no longer inferior to DALL-E-3, Midjourney, or Stable Diffusion!

As verified by my buddies, the free FLUX model now generates images that are not inferior to DALL-E-3 and MJ.
Next Controlnet and ipadapter developed after the free control of the screen composition style and so on, DALL-E-3 and MJ basically no advantage, only disadvantage.

1. Looking at the graph first, it is clear that FLUX is much more advanced in terms of ELO rating scores.

The organization, known by its acronym BFL (Black Forest Labs Black Forest Labs), is an organization that redevelops and advances advanced generative deep learning models for media such as images and video, and pushes to inspire creativity in models that break the boundaries of efficiency and diversity. Just in August, the release of FLUX.1 The Model Suite, which is a new technology that defines image detail, cue command adherence, style variety, and scene complexity for text-to-image synthesis.

To strike a balance between usability and model functionality, FLUX.1 is available in three variants: FLUX.1 [pro], FLUX.1 [dev], and FLUX.1 [schnell]:

• FLUX.1 [pro]: The best of FLUX.1, delivering state-of-the-art performance image generation with top-notch cue-following, visual quality, image detail, and output versatility. By API Access is granted by registering FLUX.1 [pro]. Alternatively FLUX.1 [pro] can be accessed via the Replicate cap (a poem) fal.ai Get. Functionality for individuals and customized solutions for businesses.
• FLUX.1 [dev]: FLUX.1 [dev] is a non-commercial, FLUX.1 [dev] evolved directly from FLUX.1 [pro], with similar raw map quality and shortcut capabilities, while being more efficient than standard models of the same size, and can be used directly in the Replicate maybe fal.ai Try it on.
• FLUX.1 [schnell]: relative to the above two, it is the fastest model and is tailored for local development and personal use.FLUX.1 [schnell] is publicly available under the Apache 2.0 license. The inference code can be found in theGitHubcap (a poem)HuggingFace's DiffusersFound in.

I believe the FLUX.1 model suite will soon enable ComfyUI integration.

2. Looking at the graph again, it is clear that FLUX.1 [pro] is the most generative, but also the most expensive.

3. BFL has released a performance comparison on its website, and FLUX.1 [pro] and [dev] outperform DALL-E 3 (HD), Midjourney v6.0, and SD3-Ultra.

The radar charts represent the comparison of each model in each of these areas [visual quality], [command compliance], [size/aspect ratio variability], and [typography and output diversity].The three FLUX.1 models were specifically fine-tuned to maintain the full output diversity of the pre-training. The advantage over the current state-of-the-art seems to be significant!But the official website sells itself, just take a look at it, in short, it does have a lot to offer.

 4. FLUX.1 All models support a wide range of aspect ratios and resolutions from 100,000 to 2,000,000 pixels.

Finally, BFL claimed that based on the FLUX.1 text-to-video modeling suite, it will launch a generative text-to-video system, SOTA, for all text-to-video scenarios, which will assist media creation and editing with high clarity, fast generation speed, accurate quality, and more. It's pretty good, we look forward to that day, it's better to come out with a free trial first, and then don't make it too expensive.

The layman's guide to getting started with AI

Help you learn how to utilize AI tools at a low cost and from a zero base.AI, like office software, is an essential skill for everyone. Mastering AI will give you an edge in your job search and half the effort in your future work and studies.

View Details>