• segs/ contains the raw data from the FOI to Irish Rail
• wavs/ contains decoded audio from the .seg files
• mp3s/ contains the decoded audio in MP3 format for wider compatibility and lower file size

Each of the above folders contains various voice packs supplied or managed by KeTech.

The most well-known are likely Phil Sayer and Celia Drummond due to their extensive coverage of the UK railway network during their time as announcers for the KeTech, Amey and Ditra customer information systems.

We're currently working to transcribe all announcements within CSV files in this repository, starting with Phil Sayer (Male1).

This is a mammoth task, and takes some time. The hope is that, with one voice transcribed, the IDs of other voices' audio snippets, or at least Celia's, will predominantly match Phil's.

You can see the current progress here.

A kind user on RailForums managed to work out a method of decoding the "proprietary" format.

On Debian/Ubuntu:

sudo apt install sox
find -iname "*.SEG" -exec bash -c "tail -c +48 '{}' | sox -t raw -b 8 -r 16000 -e a-law -X - '{}.wav'" \;

This command:

1. Iterates over every .SEG file
2. Provides the content of each file, skipping the first 48 bytes, to sox
3. sox then converts the audio into a more widely used .wav format

On Linux:

sudo apt install parallel
find "$(pwd)" -type f -name "*.wav" | parallel 'ffmpeg -y -hide_banner -loglevel error -i "{}" -q:a 0 "$(dirname "{}")/$(basename "{}" .wav).mp3"'

On Linux:

sudo apt install rename
rename 's/\.SEG|\.seg//' *
rename 's/\.WAV|\.wav//' *