Open-source music sketchpad: AI writes parts into a piano roll you can edit
Emerged finds new products worth a look, daily — what’s new today
An agent can check the sound it just made instead of taking your word for it, with no GPU and nothing leaving the machine after the first download.
earshot is a command-line tool, a Python API and an MCP server. It describes what a clip sounds like, scores how closely it matches a text description, and finds the strongest moments in a long recording. Forge writes effects as code and refines them against those scores; fit turns a recording into a synthesis recipe; 252 built-in effects render with no model and no network. MIT, Python 3.10+ and ffmpeg, 1.1GB of models on first use. One blind test of forge output came out 8 of 15.
More like this
Patch a ruined line of speech with the same moment from your other recorder
A turn-taking model that knows when a voice agent should speak
One open 9B audio model for speech recognition, understanding, TTS and editing
Edit podcasts and video by editing the transcript, entirely on your device
A Mac capture inbox for selections, dictation and screenshots, kept as files