Stability AI, a company that develops generative AI models for text and image generation, has introduced StableStudio, an open-source version of its popular DreamStudio web app. StableStudio allows users to create realistic and artistic images from text prompts, using Stability AI's state-of-the-art image generation model, Stable Diffusion.
Stable Diffusion is a generative AI model that uses a diffusion process to gradually transform random information into coherent images. Programs like stable diffusion and dreambooth can generate images of various categories, such as animals, landscapes, portraits, and abstract art, as well as images that combine multiple categories or concepts.
DreamStudio was the first consumer application that showcased the capabilities of Stable Diffusion. It was launched in August 2022 and has since attracted millions of users who have created and shared millions of images using the app. DreamStudio also features a gallery of curated images, a community forum, and a premium subscription option that offered higher resolution and faster generation.
However, Stability AI decided to release an open-source version of DreamStudio, called StableStudio, in order to promote collaborative and transparent development of the app. According to Stability AI's blog post, “We believe the best way to expand upon that impressive reach is through open, community-driven development rather than a private iteration of a closed-source product.”
StableStudio aims to be a platform for open-source collaboration and experimentation with generative AI. It offers several features and benefits for users and developers, such as:
- A new plug-in system that allows users to customize and extend the functionality of the app
- A local inference option that enables users to run Stable Diffusion on their own devices using WebGPU
- A desktop installation option that allows users to use StableStudio offline
- A chat interface that will be integrated with Stability AI's upcoming chatbot model
- A bounty program that will reward developers for improving and adding new features to the app
Stability AI's Ongoing Commitment to Open AI Development
StableStudio also demonstrates Stability AI's commitment to open-source and ethical AI development. Stability AI creates open source AI tools that anyone can access. It has AI models for various media types, such as images, words, sounds, videos and more. It also offers apps and plugins that make it easy to use its AI models. The company is the world's leading open source generative AI developer and runs on the Ezra-1 UltraCluster, the fifth most powerful supercomputer in the world.
Earlier this month, Stability announced the Stable Animation SDK in beta preview. The Stable Animation SDK enables users to easily generate 3D models of any object, character, or scene by simply describing them in words. For example, one can write “a dragon with red scales and green eyes” and get a 3D model of a dragon that matches the description. Users can also modify the models by adding or changing attributes, such as color, shape, size, texture, pose, and more.
The Stable Animation SDK is now available for beta testing for selected developers. Interested users can sign up on Stability AI's website to get early access and updates on the SDK. Users can also learn more about Stability AI's other products and services on their website or follow them on social media for the latest news and announcements.
Stability AI is at the Forefront of a Push Against Big Tech AI
We previously covered a leak of a Google internal document that showed the company's worries that open source providers like Stability AI will beat private developers such as Microsoft, OpenAI, and Google. These companies have well-known AI tools, Bing Chat (Microsoft), ChatGPT (OpenAI), and Bard (Google).
The document provides insights into the semiconductor industry, shows that Google and OpenAI are at risk of being overtaken by their open-source competitors.
The memo states, “We have no moat, and neither does OpenAI”. It also highlights the fast improvements made by open-source AI projects, which have arguably been more innovative than Google and OpenAI in recent months. The document mentions several remarkable feats by open-source AI, such as running foundation models on a Pixel 6 at 5 tokens per second, fine-tuning personalized AI on a laptop in an evening, and creating multimodal models in record time.