AI Media Skills
Media tools for AI assistants such as Codex and Claude Code.
Podcast Audio Repair
Clean up noisy or uneven speech in your podcast recordings.
- Reduce background noise.
- Even out speech levels.
- Listen to a before-and-after sample.
Requires a local AI agent, Python, and FFmpeg. The skill is free; your agent may have its own costs.
Setup instructions
- Open a working folder in a local AI agent, such as Codex or Claude Code.
- Copy the prompt below and paste it into the agent. It will check dependencies and install the skill.
- After setup, ask it to repair a 30-second sample of your recording. Listen to the comparison before processing the full file.
The prompt sets up the skill; it does not process recordings.
Install the podcast-audio-repair skill from:
https://github.com/rovidxmedia/AI-Media-Skills
Fetch the repository into a separate local folder, preserving any existing checkout.
Read its README, skills/podcast-audio-repair/SKILL.md, and helper script before running it.
Use the README's documented skill-location table to install the complete
skills/podcast-audio-repair folder for the agent I am currently using in this
workspace. Choose the project location, not a global installation. Preserve all
scripts, references, licenses, and assets. Do not overwrite an existing skill;
inspect it and explain any conflict first.
Check Python, FFmpeg, ffprobe, and the required audio filters. Set up an isolated
Python environment and install the skill's requirements.txt there. If a required
system tool is missing, explain what I need to install.
Run the installed helper with --help using that environment's Python. Confirm the
bundled model and supporting files exist. Report the skill location, Python
executable, and any missing requirements or session reload needed. Distinguish
files installed successfully from the agent actually discovering the skill.
Do not process any recordings yet.The helper processes audio locally. Your agent host has its own privacy practices. Agent setup may vary; see the full guide for supported environments.
