So compare Bild-KI from the Black Forest against Dall-E, Stable Diffusion and Midjourney

Once there is a click, it becomes a quick shave, a where you are. Behind the new Open-Source model Flux.1 is Black Forrest Labs, a small firm from Freiburg, which is now at 14 months: this is under the first old Monate alt ist. Dennoch provides the generative image with his erggebnissen in the Netz for Aufsehen and will praise his own Angabe to make a model that in the middle of the journey or the Dall-E surpass.

When the Erfolg did not come from the first generation, the Gründer Robin Rombach, Andreas Blattmann and Patrick Esser were together with the Ludwig-Maximilians-Universität Munich with Thema KI geforscht and Rombach and Blattmann present until March 2024 and the Entwicklung von Stable Diffusion improved.

Grund is a great way to make a click on Flux.1 Pro and improve. It is possible that the AI ​​image generators will take the models out of the series. Genauer has Stable Diffusion XL, Midjourney V6 and Dall-E 3. He has received the four AIs with four unused clues and are very good at improving image quality, creative settings and handles with less differences. Of course, this test can not be completed, but the fire can provide more clues. There is an end to the high quality of the different model, the man has ended up in the fall. Small spoiler: Bad does not hit Flux.

The different AI models

Flux.1 Pro

Black Forest Labs offers Flux in three different models and. Flux.1 (fast) is, however the name is displayed, the most recent images can be generated and can be a representation of the cuddly face. By finding the model Flux.1 (developer), we can use the largest model Flux.1 (pro).

For this article the web interface of Flux.1 (pro) is used in case of a failed display. By shifting the AI ​​model of the selection the credits become more valuable. The Pro-Variant of Flux costs dort pro Generierung 0.05 US-Dollar for a Dollar that can also get 20 Mal nuts with this model.

Stable diffusion V3

An error was found when finding stable diffusion in active version 3. The model was used by generating Flux, for a dollar that was started 29 times.

Developed with Stable Diffusion of Stability AI in London, a firm dies in the decaying Monaten in the Schlagzeilen geriet. After a retrospective of the CEOs followed a Stellenabbau of 10 Percent and in Great Britain the companies raise several complaints due to Urheberrechtsverwalzungen in favor. There is a reward for a billion dollars that can be used in the future, while the stability of AI will increase the money and man after a few things will do.

Midjourney V6

Wie Flux is based in Midjourney at the Forschung, and works at the Forschungsinstituut Midjourney, Inc. in San Francisco. As of 2022, a Discord server will be available in open beta. Version 6 is due out in late 2023 and will be all the texts in Bildern that you can use.

You can pay the Midjourney fee and start with the basic plan from 10 US dollars in a month. If there is a new web interface, there should be an image displayed on a Discord server that is generated.

Dall-E 3

We advised Dall-E to test the AI ​​firm OpenAI. It is the generative Bild-AI which is no longer a proprietary standard, which is integrated into ChatGPT. A text bot with the prompt is displayed and part of the instructions is a generated image.

From October 2023, Dall-E 3 will be available for the latest ChatGPT Plus subscriptions and will also allow enterprise customers to be added. At OpenAI, a fee starts at 20 dollars per month.

The Prompts and the Achievements

If you see four clues, you can test different styles and motifs. So bitten by the AIs a room ship in comic style to set, a possibly imaginary dragon and a sea, a Bleistiftzeichnung a Tanzgesellschaft and the Porträt a man, der in de kamer winkt.

In the English prompts, the messages are displayed that they will be displayed in the style, which will be a problem. If there are elements that cause sollen, we describe them. It is a fact that the errors of the construction and the color are relatively relative and somewhat comparable.

Prompt 1: Futuristic science fiction comic

Prompt: A huge, ultra-modern spaceship floats above a dystopian cityscape at sunset. The city is shrouded in mist, with glowing holograms and futuristic billboards. The spaceship is detailed, with lots of tiny lights and shiny metal plates. Below in the city, tiny people and futuristic vehicles fly between the skyscrapers. The sky is bathed in intense orange and purple hues, all rendered in a bold, dynamic comic book style with exaggerated lines and vibrant colors. (Image: Montage t3n)

Immediately noticeable, so that the Flux is in the Midjourney and Stable Diffusion, which is the Prompt. All three are Lila and Gelb and the Comic-Buch-Style come into contact with each other. If an article is written after the Flux in a comic, the others will both no longer end up in a Graphic Novel. Heraus founded in this Beispiel Dall-E with blue dyes and an insgesiges not so simple Darstellung.

Question 2: Fantasy-Drache

Prompt: A majestic dragon with glittering emerald scales rises from a crystal clear lake in the middle of an ancient, enchanted forest. Waves crash around the dragon and you can see the intricate details of its scales and their reflections in the water. The trees in the background are densely covered with leaves and magical lights shine between the leaves. A mysterious figure dressed in medieval garb stands on the shore and observes the scene. (Image: Montage t3n)

The prompt is one of the fun things you can do while testing the overall most important results. Found: Stable diffusion is not possible when generating a “mystery figure” with the image it integrates. Here is always only the dragon to be seen alone. If you have an idea, you can put the Midjourney in our best foot.

Prompt 3: Historical Bleistiftzeichnung

Prompt: A lavish ball in a 17th-century palace. The ballroom is decorated with opulent chandeliers, gilded mirrors, and richly decorated wallpaper. Noble ladies in elaborate, layered gowns and gentlemen in elegant suits and wigs dance a minuet. The intricate details of the fabrics, the facial expressions of the dancing guests, and the play of candlelight on the polished floors are depicted in a detailed pencil drawing, with fine shadows and textures that give depth to the scene. (Image: Montage t3n)

When you pick up the thread again, there is a clear difference between the Dance Hall from the 17th century. Stable diffusion generates under an oil image, focused on a slick or an oil stain from time, while others leave the blue spot in a fast way. I come across a real problem halfway through the journey all four seem to have problems with the faces of the people.

Prompt 4: Porträt eines Mannes

Prompt: A photorealistic image of a man in his mid-30s. He wears black-rimmed glasses and has short, neatly styled hair. He smiles heartily as he waves his right hand toward the viewer. The background is softly blurred to keep the focus on the man, making it look like a professional headshot. The lighting is natural, highlighting his facial features and creating a friendly, approachable atmosphere. His clothing is casual but stylish, with a simple, well-fitting shirt. (Image: Montage t3n)

The larger photos are so cute that they quickly make a portrait photo of a man with glasses. Dall-E is now working here as a comic strip and not photorealistic. If Flux were a man, he was convinced that an AI image would recognize, while halfway through the journey the journey is so good, that the image of a real photo no longer exists. A stable diffusion said that the hand in 2024 can still be a problem for an intelligent intelligence.

Decisions and (disappeared) borders

If you give a description, this test can now end a first time with the failure and the undoing of the generative AI models. If the first Blick reaches the Flux stream, the AI ​​is not so long on the way with the Mitbewerber. Should we put us on a ranking, Flux cuts really well at the four suggested Beispielen with the second place. Midjourney is part of a photorealistic image of people and makes a large bandwidth.

There are a number of things that are separated from the flow and the rest, but it is not possible to go a corridor and end the bets of KI. Während zum Beispiel Dall-E sich refuse Photos of prominent people who pose, how Flux has no problem. A man who sees the web interface of the Fal.ai security tolerance of Flux.1 (pro) can then use an image with weapons or other nackten generated in the second generation. Donald Trump is his father a soul or Putin? Erstellt Flux ohne zu murren!

That also applies, that the same Photos are made for everything at X the Runde-machine, denn-zwischen X and Black Forest Labs give a collaboration. Paying X-Subscriptions:innen of the Flux name in the social network, an image for general messages. A person who adheres to the obligation will be surprised or have the safety tolerance high on the banner.