✦
Summary
Zoo by Replicate is an interactive playground for testing and comparing generative image models. Built by Replicate, it lets artists, developers and researchers experiment with text-to-image models and understand their strengths and weaknesses by comparing outputs from the same prompt side by side.
Key Features
Side-by-side comparison of text-to-image models
Supports Stable Diffusion 1.5, 2.1, SDXL, Kandinsky-2 and DALL-E
Parameter manipulation for output control
X/Y plots to explore parameter effects
ControlNet support
Open-source, cloneable and runnable locally from GitHub
PostgreSQL and Supabase integration
Memorie feature for saving generations
About Zoo by Replicate
Users enter a text prompt and generate images from multiple models at once, including Stable Diffusion 1.5, 2.1 and SDXL, Kandinsky-2 and DALL-E, adjusting parameters such as resolution and style, and using X/Y plots and ControlNet to explore how settings change the result. The platform is open source, so it can be cloned from GitHub and run locally, with your own models added and PostgreSQL or Supabase used for data management.
Zoo is aimed at anyone exploring generative image models — from artists choosing a model for a project to developers and researchers testing new techniques — and it is free to use and open source.