Black Forest Labs: Image editing via AI prompt
The artificial intelligence "Flux.1 Kontext" from Freiburg-based AI start-up Black Forest Labs analyzes the image content to process it via prompt input.
The artificial intelligence "Flux.1 Context" analyzes the image content and places motifs in a new context after prompt input.
Generative artificial intelligence is commonly understood to mean image generators that show an astronaut on a unicorn after text input or convert uploaded photos into drawn illustrations. The new Flux.1 model family from AI start-up Black Forest Labs in Freiburg im Breisgau aims to go beyond this. It should be able to process images using text instructions. Company founders Robin Rombach, Andreas Blattmann and Axel Sauer were able to gain experience for their AI at the London-based company Stability.AI in addition to studying in Heidelberg and Munich.
What Flux.1 does in context
According to the company's website, Flux.1 Context identifies and preserves the characters in the image and can thus place figures or objects in a different context. For example, to stage the CEO at the lectern in a new image as a skateboarder on the ramp. In the simplest case, the AI adds objects, such as a VR headset, or removes objects, such as a hat or glasses.
Some functions are already familiar from other AI models such as Adobe Firefly: Flux.1 can use uploaded images as a style reference to create an image of an entirely different subject in the same style from a charcoal drawing of an apple, for example.
With Flux.1 Context, it should be possible to work iteratively, i.e. to add further instructions to a project that has already been started to build on a previous result.
Videos by heise
Three variants of the model family
Black Forest Labs is introducing three variants of its Flux.1 family. The premium model Flux.1 Context [max] offers the maximum possible performance at high speed. It should understand prompts better than previous models and be able to generate typography.
The Flux.1 Context [pro] model focuses on fast processing. It is ideal for brainstorming or iterative processing with many work steps. It should also be suitable for local image processing. The manufacturer also plans to offer the Flux.1 Context [dev] model for research purposes in the future.
Try out the AI for yourself
Flux.1 Kontext can be tried out in the "Playground". Users receive 200 credits, which are sufficient for around 12 image generations. The English-language interface offers the editing options Generate, Edit, Fill and Expand.
The company provides detailed prices for the entire model family and for editing individual images. It is aimed at major customers and developers. Deutsche Telekom, for example, cooperates with Black Forest Labs.
(akr)
