Flux 1.1 Pro is six times faster with improved image quality, prompt adherence, and diversity than its predecessor.
Black Forest Labs has just released its most advanced, efficient, and powerful open-weights image model to date: Flux 1.1 Pro. This new image model offers a significant performance boost, delivering six times faster generation speeds than its predecessor while also improving image quality, prompt adherence, and overall diversity.
For businesses and indie developers like myself, who rely on Flux models for image generation and fine-tuning capabilities in our products, this update is a huge deal.
For those unfamiliar with Flux image models, they are a suite of open-weights text-to-image models that define a new state-of-the-art (SOTA) in image detail, prompt adherence, style diversity, and scene complexity for text-to-image synthesis.
Flux models come in three variants:
All public Flux 1 models employ a mix of multimodal and parallel diffusion transformer blocks and are equipped with a staggering 12 billion parameters. These models surpass earlier diffusion models by using flow matching, a method for training generative models that improves both speed and quality.
The newly released Flux 1.1 Pro brings several key improvements over its predecessor. As mentioned earlier, it provides six times faster generation speeds. Hereâs a breakdown of whatâs new:
Notice how Flux 1.1 Pro surpasses other popular image generators like Midjourney and OpenAIâs Dall-E 3.
Aside from the faster generation speed and improved image quality of Flux 1.1, Black Forest Labs also updated the Flux 1.0 Pro model to generate the same output as before, but two times faster.
In the examples below, I will be using Flux 1.1 Pro on Replicate.
Prompt: black forest gateau cake spelling out the words âFLUX 1.1 Proâ, tasty, food photography
The image quality is significantly sharper than what I achieved with Flux 1.0 Pro, and the generation speed has noticeably improved. While I didnât measure the exact speed increase, the output felt much quicker on Replicate.
Prompt: Polaroid photo with VSCO filter, 1990, gorgeous woman, night, flash photo, blonde, cute, young face, beautiful shadows, tropical plants, urban clothing, inside an apartment, DSLR, holding a sign written in ballpoint pen on a notebook saying âThis is an example long text to test out Flux 1.1 Proâ
In this example, I wanted to test Flux 1.1 Proâs ability to render long text. The result was legible, with a pleasing font style and color on the paper. However, the AI did make an error by repeating the word âtestâ twice.
Also, I intentionally removed the words âof aâ between âThis is an example [of a] long textâŠâ to see if the AI wonât correct the grammar on its own.
Prompt: boring snapchat photo
The photorealism in this example was mind-blowing. Without pixel peeping, it could easily pass as a real photo.
As some of you who follow my blog on Medium and Substack know, I develop products and promote them on my websites. Just a few weeks ago, I launched Flux Labs AI, a platform that allows users to fine-tune or train Flux image models with their own images.
Flux Labs AI offers image generation capabilities with all three Flux models. The introduction of the faster and more capable Flux 1.1 Pro will greatly enhance the platformâs usability. While the API is not yet updated to support the latest model, I plan to work on that over the weekend to ensure smooth integration.
I am super excited about it, and I hope you are too :)
In case you havenât heard about Flux Labs AI yet, feel free to check out the app here:
Flux Labs
Train Flux image models for your custom images.www.fluxlabs.ai
You can also read my previous article detailing what Flux Labs AI is all about and how to generate images or train models:
If youâve already tried it, Iâd love to hear your thoughts!
Flux 1.1 Pro is now available on several platforms with API access, including:
Black Forest Labs also offers its own API access, which provides the following benefits:
Hereâs an example API call to generate an image with the Flux 1.1 Pro.
const request = require('request');
const options = {
method: 'POST',
url: 'https://api.bfl.ml/v1/flux-pro-1.1',
headers: {'Content-Type': 'application/json', 'X-Key': 'YOUR_TOKEN'},
body: {
prompt: 'ein fantastisches bild',
width: 1024,
height: 768,
prompt_upsampling: false,
seed: 42,
safety_tolerance: 2
},
json: true
};
request(options, function (error, response, body) {
if (error) throw new Error(error);
console.log(body);
});
Hereâs an example endpoint for retrieving a generation task result.
const request = require('request');
const options = {method: 'GET', url: 'https://api.bfl.ml/v1/get_result', qs: {id: ''}};
request(options, function (error, response, body) {
if (error) throw new Error(error);
console.log(body);
});
Flux 1.1 Pro with fast ultra high-resolution generation is coming soon to the API. Generate up to 2k images without sacrificing any of the prompts following.
The release of Flux 1.1 Pro took me by surprise, especially since it comes only a few weeks after the initial launch of Flux 1.0. In terms of speed and quality, Flux 1.1 Pro delivers noticeable improvements. However, I did encounter some minor issues with long text prompts and short prompts where the output quality wasnât particularly good.
Another exciting aspect of this release is the API access. Black Forest Labs offers competitive pricing, which makes it a tempting option for those looking to switch from providers like Replicate or Together AI. Iâm eager to see if this will benefit my own Flux Labs AI project as I consider changing my API routes.
Itâs been a crazy week in the AI world with lots of big updates from the big players like OpenAI, Runway, Kling, and now Black Forest Labs. So, what are your thoughts on Flux 1.1 Pro? Let me know in the comments.
Make sure to subscribe to stay updated on the latest developments in the AI space.
â
Software engineer, writer, solopreneur