AGENT USE CASE1:21Let sound design lead the scene
Building many scene variants with sound first, the creator says the result looks less like AI slop and renders fast.
Claude Code♥ 988/
Get Friday PicksSubmitearshot: give your coding agent ears for sound
A local tool and MCP server, with two skills, that lets Claude Code or Codex judge audio it cannot hear: describe a sound, rank candidates against a description, find the best moment in a recording, check loudness and sync, and critique a whole video's soundtrack. It can also render sound effects from a bank made in code for code-made films.
In short: A local tool and MCP server, with two skills, that lets Claude Code or Codex judge audio it cannot hear: describe a sound, rank candidates against a description, find the best moment in a recording, check loudness and sync, and critique a whole video's soundtrack. It can also render sound effects from a bank made in code for code-made films. It is a free skill for people who use Claude Code or Codex and want help with sound & music.
Checked on · Added
Works with
| Agent / tool | Access / cost | Link |
|---|---|---|
| Claude Code | Paid Claude plan or API | claude.com/claude-code |
| Codex | ChatGPT plan or API | openai.com/codex |
| GitHub repo | Open source · ★ 3.1k | Leonxlnx/earshot |
Image: from the GitHub README, shown with credit.
Images from the project README on GitHub.
Try it in 60 seconds
git clone https://github.com/Leonxlnx/earshot && cd earshot python -m venv .venv && .venv/bin/python -m pip install torch --index-url https://download.pytorch.org/whl/cpu .venv/bin/python -m pip install -e ".[mcp]"
Install command copied from the project README. Check the README for the latest version.
What it does
A local tool and MCP server, with two skills, that lets Claude Code or Codex judge audio it cannot hear: describe a sound, rank candidates against a description, find the best moment in a recording, check loudness and sync, and critique a whole video's soundtrack. It can also render sound effects from a bank made in code for code-made films.
Agents can write the code for a film, demo or game but cannot hear it. earshot turns audio into readings an agent can use: what a file sounds like in words, how well it matches a description, the best moment in a long…
Short excerpt from the README of Leonxlnx/earshot. Read the full README on GitHub.
Inside the repo
What it doesMake sounds: earshot forgeearshot fit: audio to codeSound for films: bank, sfx, kitsGalleryInstallMCP server for Claude CodePythonAgent skillsHow good is it?LimitsRepo map
A prompt to try
Use earshot to check whether the sound effects in demo.mp4 land on the cuts, and fix the ones that don't.
Use earshot to check whether the sound effects in demo.mp4 land on the cuts, and fix the ones that don't.
Written by us so you can reuse it. Replace each [bracket] with your details.
Step by step
- Open the GitHub repo and watch the 22 s launch film (all its sounds were made with earshot).
- Clone it, make a virtual env and install it with the mcp extra (see the README's Install section).
- Add it as an MCP server (claude mcp add earshot ... -m earshot mcp) and copy skills/earshot into ~/.claude/skills/.
- Ask your agent to pick the best whoosh for a cut, or to critique your video's audio.
Watch out for
The first use downloads two open audio models (about 1.1 GB) into the repo folder, and each CLI call takes 20-40 s to load them, so use the MCP server for more than a few calls. The optional llm command sends audio to Gemini with your own key; everything else stays on your computer. The real-sound skill uses licensed recordings you bring yourself.
Docs, demos & materials
Common questions
What does earshot do?
A local tool and MCP server, with two skills, that lets Claude Code or Codex judge audio it cannot hear: describe a sound, rank candidates against a description, find the best moment in a recording, check loudness and sync, and critique a whole video's soundtrack. It can also render sound effects from a bank made in code for code-made films.
Which AI agents can use earshot?
It works with Claude Code and Codex. The code is on GitHub at Leonxlnx/earshot.
Is earshot free?
Free & open source (MIT). Needs Python 3.10+ and ffmpeg; runs on CPU, no GPU or API key.
How do I install earshot?
Run the install command from the README: “git clone https://github.com/Leonxlnx/earshot && cd earshot”. Check the README on GitHub for the latest version, then follow the steps on this page.
Is earshot safe to install?
We read 237 instruction and code files in the repo and found no red flags. That is not a guarantee, so review anything that runs commands on your machine.
Sources & credit
All credit to the original authors. We rewrite in our own words and link back. Something wrong? Report it.
Related skills and use cases in Sound & Music
- Let sound design lead the scene Use case
- Music video coded in Codex with Remotion + Suno Use case
- AudioCraft: MusicGen + AudioGen SFX Skill
- ACE-Step: free local music generation Skill
- Make a track in Ableton Live by chatting Use case
- MMAudio: auto-foley for silent AI clips Skill
- ElevenLabs official agent skills Skill
- Agentic sound design inside the editor Use case
Related
AGENT USE CASE1:21Claude Code♥ 988
AGENT USE CASE1:30Music video coded in Codex with Remotion + Suno
A complete music video built only with code (Remotion via Codex), with the song made in Suno.
Codex★ 64.3k
SKILLAudioCraft: MusicGen + AudioGen SFX
Meta's open library for local music and sound-effect generation.
Claude Code +1★ 23.7k
SKILLACE-Step: free local music generation
Open-source music model for royalty-free beds you generate on your own machine.
Claude Code +1★ 4.9k
AGENT USE CASEMake a track in Ableton Live by chatting
An MCP server lets the agent drive Ableton: drums, bass, melody and a rough mix from chat.
Codex★ 229
SKILLMMAudio: auto-foley for silent AI clips
Generates synced sound effects from a silent video, ideal for Seedance clips.
Claude Code +1★ 2.3kFree newsletter
Friday Picks
New AI tools you can use right away, in plain words. Every Friday.