diff --git a/tools/audio-offset/README.md b/tools/audio-offset/README.md
new file mode 100644
index 0000000..01aa0b8
--- /dev/null
+++ b/tools/audio-offset/README.md
@@ -0,0 +1,63 @@
+# Audio Offset Tool
+
+Finds the time offset between a local audio file and a YouTube video's audio using Chromaprint fingerprint matching. Generates stereo comparison WAV files for ear verification.
+
+## Prerequisites
+
+### ffmpeg
+
+```bash
+# macOS
+brew install ffmpeg
+
+# Ubuntu/Debian
+sudo apt install ffmpeg
+
+# Windows (via scoop)
+scoop install ffmpeg
+```
+
+### fpcalc (Chromaprint)
+
+```bash
+# macOS
+brew install chromaprint
+
+# Ubuntu/Debian
+sudo apt install libchromaprint-tools
+
+# Windows
+# Download from https://acoustid.org/chromaprint — extract and add to PATH
+```
+
+### Python dependencies
+
+```bash
+pip install -r requirements.txt
+```
+
+## Usage
+
+```bash
+# Basic usage
+python find_offset.py ../../public/songs/get-lucky/audio.ogg 5NV6Rdv1a3I
+
+# Video IDs starting with '-'
+python find_offset.py ../../public/songs/meet-me-halfway/audio.ogg --yt=-GY8MhHMxLM
+
+# Custom output directory
+python find_offset.py ../../public/songs/get-lucky/audio.ogg 5NV6Rdv1a3I --out-dir ./my-compare
+
+# JSON output (for scripting)
+python find_offset.py ../../public/songs/get-lucky/audio.ogg 5NV6Rdv1a3I --json
+```
+
+## Output
+
+The tool prints candidate offsets ranked by fingerprint similarity and writes stereo WAV files to the output directory.
+
+Listen to each WAV with headphones:
+- **Left ear** = local audio
+- **Right ear** = YouTube audio (shifted by candidate offset)
+
+The file where both ears are in sync has the correct offset.