Find a Replicate model with AI
Describe the inputs you have and the output you need, then get a shortlist based on current model documentation rather than a familiar name.
Find a Replicate model with AI by describing the material you already have. A photo of a chair plus a fabric sample calls for something different from a prompt that asks for a new chair from scratch. Both are image tasks, but only one requires reference inputs.
The find-models workflow searches current listings and collections, then inspects individual model schemas. It can leave you with a shortlist and an explanation of what each candidate accepts, without generating any output.
What it does
- Search for a capability rather than a model name.
- Check whether a candidate accepts a reference image.
- Inspect the output format before planning an integration.
- Distinguish an official model from a community version.
- Save the owner, model name and relevant version for testing.
Install and use
- Give the agent a task, required inputs and intended output. API search needs appropriate Replicate access; public model pages can supply documentation. Paid predictions are a separate decision.
- Try: “Find models that accept a chair photo and a fabric reference. List the inputs for each candidate, but do not run them.”
- Read the shortlist against your requirements. Flag a feature that appears in a description but not in the documented input schema.
FAQ
Source
Made by Replicate. Original skill: find-models, published on GitHub under the Apache-2.0 licence.
View on GitHub