| title | Podcast Mic Ducking |
|---|---|
| emoji | 🎙️ |
| colorFrom | blue |
| colorTo | purple |
| sdk | streamlit |
| sdk_version | 1.55.0 |
| app_file | streamlit_app.py |
| pinned | false |
Clean up your two-mic podcast recordings automatically.
When you record a podcast with two people in the same room, each microphone picks up both voices. The person sitting farther away sounds echoey or distant on the other mic. This app fixes that.
You give it two audio files — one from each microphone. The app figures out who's talking at each moment and turns down the mic that isn't being used. This removes the echo and background bleed, so each speaker sounds clean and close.
It also:
- Evens out the volume so quiet words and loud words are closer to the same level
- Sets the overall loudness to the standard that podcast apps like Spotify and Apple Podcasts expect
- Prevents distortion by making sure nothing is too loud
Use the web version — no installation needed:
Upload your two files, click Process, and download the results.
Running locally is faster (especially for long recordings) and has no file-size limits. Here's how:
If you don't already have Python, download it from python.org. Version 3.10 or newer works. During installation on Windows, check the box that says "Add Python to PATH".
Click the green Code button at the top of this page, then Download ZIP. Unzip it somewhere you'll remember (like your Desktop).
Or if you're comfortable with the terminal:
git clone https://github.com/MikeyJarrell/ducking.git
cd ducking
Open a terminal (Mac: Terminal app, Windows: Command Prompt or PowerShell) and navigate to the folder you just downloaded. Then run:
pip install streamlit torch numpy scipy silero-vad soundfile
This downloads the libraries the app needs. It may take a few minutes the first time (PyTorch is a large download).
From the same terminal, run:
streamlit run streamlit_app.py
A browser window will open with the app. Upload your two mic files, click Process, and download the cleaned-up versions.
There's also a simpler desktop version with a basic window interface (no browser needed):
pip install torch numpy scipy silero-vad
python ducking_app.py
If you're on a Mac and want to launch Ducking from Spotlight or the Applications folder like any other app, run:
chmod +x make-app.sh
./make-app.sh
This builds a standalone Ducking.app using py2app and installs it to /Applications. The bundle embeds Python, PyTorch, and all dependencies — no system Python needed at runtime — so it's ~600 MB but self-contained.
Requires python.org Python 3.12 (from python.org/downloads) — anaconda's Python is not a framework build and won't work with py2app.
- WAV (recommended — lossless, best quality)
- MP3, FLAC, OGG (the web version accepts these too)
Output is always WAV.
The app uses a machine-learning model called Silero VAD to detect when someone is speaking. It then compares the volume levels between the two mics to figure out which speaker is talking. The mic that isn't being used gets turned down (not all the way to silence — a little room noise is kept so it sounds natural). Smooth volume fades prevent any clicking or popping at the transitions.