Skip to content 
Less than a week ago FLUX.1 was released, leaving followers of artificial intelligence perplexed with the quality of its results
It has appeared almost out of nowhere and without warning, and has revolutionized the world of artificial intelligence for image generation with its absolutely faithful results and a quality worthy of the most professional models.
And that is exactly what is most striking, that everything we see next It is the result of a model that was just released a few days ago., in continuous development and a version 1.0 that makes us wonder: If it works like this now, how far will it go?
The origin of FLUX.1
The beginning of everything is—unexpectedly—a company from here in Europe, specifically a German one. Black Forest Labs was founded by a team of leading AI researchers and experts, many of whom previously worked at Stability AI, the company behind Stable Diffusion.
The startup It was officially launched on August 1, 2024 —is just born as of these lines—, with the aim of redefining generative artificial intelligence and establishing new standards in the creation of generative media, such as images and videos —coming soon—.
The company received a round of initial financing of 31 million dollars, led by Andreessen Horowitz, and is backed by leading industry figures such as Garry Tan, CEO of Y Combinator, and Timo Aila, principal researcher at Nvidia.
This impressive round of financing will allow Black Forest Labs expand its capabilities and bring their innovations to a global audience. First in pictures as we will see today and later, in video format.
What is FLUX.1
In a very summary way, FLUX.1 is the first version of this imaging model, a direct competitor to Dall-e or in particular Midjourney. But it has a couple of peculiarities that are worth highlighting.
First of all, he is absurdly good at understanding prompts complex and long, with several lines of requests. You can mix different environments within the same image, which is complicated in other similar models.
Secondly, he is very good at adding small details – whenever we ask him – to the images we generate. In this way it is able to understand even the smallest thing that we want to add to our creation.
And on the other hand, he writes like no other. It is extraordinarily coherent in writing text, it has small errors at the moment, but to have been accessible to the public for less than a week it is incredible. Great things await you.
How to use FLUX.1
I am going to explain to you in a simple and quick way how to use FLIX.1 completely free —at least for now— to start creating images in less than 5 minutes.
Step 1: Enter your website
To start we have to go to Black Forest Labs website, since FLUX as such at the moment does not have its own website or application. Midjourney recently created its own page, but with the model at hand we will have to take a detour.
Once on its website, as soon as you enter you have to click on the "Get FLUX" button to take us to the image generation model section.
Step 2: Choose the model
At this point we will see a page where we can choose the model we want to use from the next 3 that are proposed to us. Let's start by discarding the first and the last for now and focus on the DEV.
Model PRO. This is designed to use their API and is paid, and the model schnell It is the simplest and ready to install on your computer. We are going to proceed to use it online and also totally free for us, although perhaps with some limitations.
The model to choose is the FLUX.1 [dev]
As you can see, it can be accessed directly from the GitHub repository, but if you want something simpler and more accessible we can access the model on other platforms, for which I am going to tell you exactly how to do it.
hugging face
From this link that I leave you here You can directly access the model and start writing in the section that says «Enter your prompt» and immediately hit run to run it and see how the image looks.
Replicate
With the replicate website It happens exactly the same, with the difference that you have to register through GitHub to use it, so at the moment I have not tried it on my own, since the previous option allows you to do it without registration.
Fal.ai
This here is the last option. The only problem I have found is the limitations to generate images, and that, although it is cheap - 40 requests for each $1 - you have to pay after about 5 executions.
What do I expect from FLUX.1
For me it is a revolution, both for the ability to understand extraordinarily long prompts, have infinite details and know how to write more or less correctly. I think we will have to follow it closely.
[…] happened with the images. From those first designs with the first version of Dall-e to what Flux is capable of doing today, there is a leap in quality […]