With “StableStudio”, the AI start-up Stability AI has presented an open-source variant of its text-based image generator DreamStudio. Released under an MIT license on Github, StableStudio, like the commercial DreamStudio, serves as an interface for Stability AI’s image creation models.
The company now wants to involve the developer community in the further development of the platform. Millions of images have already been created with DreamStudio, Stability AI said on Wednesday: “We believe the best way to increase this impressive reach is through open, community-driven development.”
Development with the community
Stability AI is moving all further development to a public repository, it said. The start-up asks for wishes and ideas and is also planning a reward system for developers who contribute improvements and new functions.
Founded in 2020, Stability AI had previously released the AI generator Stable Diffusion as open source. Later, DreamStudio took over the role of the main interface for the AI models. Most recently, Stability AI introduced the SDXL image creation model and released two Large Language Models (LLM) under a Creative Commons license.
If can write
Stability AI is also behind the AI ”If” developed by DeepFloyd. “If” is intended to address a central weakness of AI-supported image generators: the artificial intelligences cannot write. “If” should now also be able to create photorealistic images with lettering – such as “Hotel” on a building – and are therefore also suitable for graphic design.
Stability AI is also the focus of the debate about AI-supported image generators and copyright. The US picture agency Getty Images has sued the start-up in Great Britain. The accusation: The AI unlawfully processed Getty images during training and thus violated copyright millions of times. Lawsuits against Stability AI and Midjourney are also pending in the United States.
(vbr)
To home page