The open-source, CPU-only toolkit analyzes soundtracks, critiques launch videos and lets agents generate sound effects, though its blind-test evidence remains small.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
Primary source: [X](https://x.com/LexnLin/status/2110878517197975717)
Why it matters #
Earshot targets a gap in agent-made media: agents can generate audio, but need ways to inspect timing, compare sounds and iterate. Its own small blind test and model-specific limits show how far automated scoring remains from human judgment.
Leon Lin (@LexnLin) released Earshot, an open-source toolkit that lets AI coding agents analyze audio, critique launch videos and generate sound effects from code. Lin introduced the project in a five-post thread on X on October 15th, describing a familiar production gap: agents can assemble a video, but struggle to tell whether its sound effects feel cheap or out of place.