Replicate

Meet Replicate

Replicate runs AI models through an API on managed infrastructure. Its catalog covers tasks such as images, video, speech, and music, with each model defining the inputs it accepts and the result it returns. Python and JavaScript examples show how to request a prediction from application code.

Developers can fine-tune supported models or deploy their own. The hosted execution lets an app request results without operating that model's infrastructure itself. Comparing candidate models involves reviewing their outputs, required inputs, and runtime behavior. The deployment and selected model determine how a request executes and which settings are available to the application.