earshot
Lets coding agents hear audio, score it, and make sounds.
- AI tool
- Free
Pricing
Free and open source. License: MIT.
Overview
earshot is a CPU-only tool that gives coding agents a way to listen. It runs as a command-line tool, a Python API or an MCP server, and it can describe what an audio file sounds like, score how well a sound fits a written description, and find the best moments in a long recording. It can also check a launch video's sound against a profile built from professional films and list timestamped fixes. A forge command lets an agent write a sound as code, read how it scores, and revise it.
Key features
- Describes a sound in words with calibrated tags and measurements
- Scores a file against a text description and ranks candidates
- Locates the strongest moments within lengthy recordings
- Critiques a launch video against a reference profile
- Forge: an agent builds a sound from code and refines it by ear
- Available as a CLI, Python API and MCP server; an optional hosted-model critique needs your own key and consent
Use cases
Alternatives
Last updated 2026-10-17