This Python script converts TRS subtitle files to the simpler and more common WebVTT format. TRS subtitles files are XML files produced by the old transcription software Transcriber and TranscriberAG.
The only other script I found to convert TRS files is geomedialab/convert-trs-srt. However, next to an easily fixable bug when encountering specific timecodes, it loses most temporal information when processing long segments by a single speaker. I thus found it easier to write my own script and skip the intermediate conversion from SubRip to WebVTT, which can be done, e.g., using nwoltman/srt-to-vtt-converter.
The script does not handle the named entity annotations described in the TranscriberAG manual, and comments are currently removed.
Python 3.6+
There are no other dependencies unless you want to replace the vulnerable standard XML library with defusedxml.
python convert.py [-h] [-o PATH] [-l LANG] [-s] [-n] INPUTFILE
Unless an output path is specified using -o or --output, the result will be written to stdout.
The -l or --language argument adds a language header to the file.
When using -s or --speakers, each spoken line is prefixed with <v SPEAKER>, if this information is annotated.
When using -n or --noise, noise events such as (laughter), (silence), or (unintelligible) are preserved.
Input TRS file example.trs created with Transcriber:
<?xml version="1.0" encoding="ISO-8859-1"?>
<!DOCTYPE Trans SYSTEM "trans-14.dtd">
<Trans scribe="(unknown)" audio_filename="18061-006" version="1" version_date="020301">
<Speakers>
<Speaker id="spk1" name="interviewer" check="no" dialect="native" accent="" scope="local"/>
<Speaker id="spk2" name="interviewee" check="no" dialect="native" accent="" scope="local"/>
</Speakers>
<Episode>
<Section type="report" startTime="0" endTime="902.624">
<Turn speaker="spk1" startTime="0" endTime="15.665">
<Sync time="0"/>
<Sync time="9.037"/>
survivor is Abraham [Bommer] the date is August fourteenth nineteen ninety six
</Turn>
<Turn speaker="spk2" startTime="15.665" endTime="34.685">
<Sync time="15.665"/>
you know something I forgot to mention is about
<Sync time="20.3"/>
a man by the name of Captain [Jello]
<Sync time="23.99"/>
he was almost the head man from the
<Event desc="EE-HESITATION" type="noise" extent="instantaneous"/>
Czechoslovakia army
<Sync time="28.385"/>
he came over there according my knowledge is
<Sync time="32.56"/>
with his wife because his wife was Jewish
</Turn>
[...]
Simple WebVTT output: python convert.py example.trs
WEBVTT
00:00:09.037 --> 00:00:15.665
survivor is Abraham [Bommer] the date is August fourteenth nineteen ninety six
00:00:15.665 --> 00:00:20.300
you know something I forgot to mention is about
00:00:20.300 --> 00:00:23.990
a man by the name of Captain [Jello]
00:00:23.990 --> 00:00:28.385
he was almost the head man from the Czechoslovakia army
00:00:28.385 --> 00:00:32.560
he came over there according my knowledge is
00:00:32.560 --> 00:00:34.685
with his wife because his wife was Jewish
[...]
WebVTT output with language, speaker, and noise info: python convert.py example.trs --language en --speaker --noise
WEBVTT
Language: en
00:00:09.037 --> 00:00:15.665
<v interviewer>survivor is Abraham [Bommer] the date is August fourteenth nineteen ninety six
00:00:15.665 --> 00:00:20.300
<v interviewee>you know something I forgot to mention is about
00:00:20.300 --> 00:00:23.990
<v interviewee>a man by the name of Captain [Jello]
00:00:23.990 --> 00:00:28.385
<v interviewee>he was almost the head man from the <i>(ee-hesitation)</i> Czechoslovakia army
00:00:28.385 --> 00:00:32.560
<v interviewee>he came over there according my knowledge is
00:00:32.560 --> 00:00:34.685
<v interviewee>with his wife because his wife was Jewish
[...]
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent