Delay-Check – Open-source tool to calculate audio delay between two files

[sicuskyle]

I've been working on a small Python tool called delay-check that calculates the time offset (delay) between two audio/video files, ideal for synchronization tasks (e.g. matching a dubbed track to an original).

How it works:

1. Extracts audio fingerprints using MFCC coefficients
2. Applies cross-correlation to find the optimal offset
3. Validates the delay across multiple segments (25%, 50%, 75%, end) for consistency
4. Reports delay, correlation score, confidence level, and time drift

Key features:
- Supports video containers (MKV, MP4, MKA) and many audio formats (WAV, AC3, AAC, DTS, FLAC, MP3, etc.)
- Multi-track selection (interactive if multiple audio tracks are found)
- Fully configurable via JSON (analysis time, sample rate, MFCC params, etc.)
- Detailed logging
- Confidence scoring with drift detection to catch VFR or different-source issues

Examples:
# Compare two audio files directly

delay-check original.ac3 dubbed.ac3
delay-check original.aac dubbed.aac
delay-check original.wav dubbed.wav

# Extract audio track from a video and compare against an audio file

delay-check movie.mkv dubbed.ac3

# Compare two video files (will use their first audio track)

delay-check original.mkv dubbed.mkv

Repo: https://github.com/sicuskyle/delay-check

I'd love for people to test it with their own files and report any issues or suggestions.

[aramkolt]

Love the concept! How exactly does it figure out what the delay actually is though? Like does it somehow see if lips appear to move with speech or other triggers in the audio, or I saw you mentioned dubs - perhaps it only works with videos that already have audio that'd be the same between them such as music or sound effects?

I'm still not clear on what the best way to add audio to say something captured with VHS-decode if the audio is being captured by traditional means.