Zoo is a playground for generating photo-realistic images from text inputs using various text-to-image AI models. The tool employs a variety of latent text-to-image diffusion models, including STABILITY-AISTABLE-DIFFUSION 1.5 and STABILITY-AISTABLE-DIFFUSION 2.1, capable of generating images based on natural language descriptions.
Additionally, Zoo incorporates AI-FOREVERKANDINSKY-2, a text2img model trained on internal and LAION HighRes datasets, and OpenAI’s DALL-E, another AI system that can create realistic images and art representations from natural language descriptions.
Users of Zoo can input any text, such as “a tilt shift photo of fish tonalism by Ugo Nespolo”, to generate a corresponding image. The tool runs on a PostgreSQL database and file storage provided by Supabase, and is an open-source repository on GitHub.
The tool is powered by Replicate, a company specialized in providing infrastructure for AI and machine learning projects.Zoo is a valuable resource for researchers and developers exploring text-to-image AI models and their applications.
With its range of models available and open-source nature, Zoo offers an accessible and collaborative space for those interested in exploring the latest advancements in computer vision AI.