audd-recognize

Shazam-like audio recognition via AudD API. Identifies songs, music, and some vocal content by audio fingerprint.

When to use

• User sends an audio clip / voice note / video and asks "what is this song / nasheed / tune / recitation?"
• User wants to identify background music, melodies, or commercial tracks
• For Quran recitations try transcription+reasoning first (usually more accurate); use AudD as fallback

Capabilities vs limits

• ✅ Commercial music (global + Arabic), humming/singing (sometimes), TV themes
• ⚠️ Modern nasheeds / obscure recitations: hit-or-miss
• ❌ Pure ambient tones with no fingerprint match

Usage

Script: /root/.openclaw/workspace/skills/audd-recognize/recognize.sh <audio_file>

Uses AUDD_API_KEY from env if set. Without a key AudD allows a very limited free tier (useful for testing, but rate-limited).

Returns JSON with: title, artist, album, release_date, plus links to Apple Music / Spotify / Deezer when matched.

Setup a paid key (optional)

1. Register at https://dashboard.audd.io/ 2. Add key to shell env: export AUDD_API_KEY=... (persist in /root/.bashrc or secrets)

Example

bash /root/.openclaw/workspace/skills/audd-recognize/recognize.sh /path/to/clip.ogg