Non MidiCo/Kareoke Format?

[ben-umble]

Hi,

Currently in the process of finding a way to generate .lrc files for some obscure music I'm hoarding. This project is great! However, my use case isn't karaoke but rather audio playing.

I'm wondering if this could be made to support a non-specific karaoke format for use with media servers or if it'd be out of scope for this project?

As an example, this is the output as-is:

[re:MidiCo]
[00:00.000]1:Here's 
[00:11.060]1:the 
[00:11.199]1:story 
[00:11.519]1:of 
[00:11.779]1:a 
[00:11.939]1:sinner, 
...

I'm new to this so I looked this up and the 1:{word} format looks to be specific for karaoke (as in singer 1). My audio player however lists each word and includes the 1: format.

This is more like what I'm looking for however:

[00:10.00]<10.00>Here's <11.06>the <11.19>story <11.51>of <11.77>a <11.93>sinner
[00:13.00]<13.06>Who <13.33>used <13.83>to <14.06>be <14.20>a <14.33>winner

The word-by-word timestamps may be a bit too much so it could as easily be:

[00:10.00]Here's the story of a sinner  
[00:13.00]Who used to be a winner  

I'm a software engineer so can help out with this but if it's out of scope for this project, I'll fork - with your permission - and make some changes there instead 👍

Thanks!

[beveradb]

Hi Ben,

Thanks for reaching out!
I'd love to help make it work for you, that seems like a very reasonable use case to me.

Have you tried the ASS (subtitle format) output already generated by this? That should play in most media players, although the positioning of the lyrics may need to be tweaked depending on where on the screen you're hoping for the lyrics to be displayed.

Alternately, the (Lyrics Corrections).json JSON output file generated by this has a corrected_segments key which could very easily be used to generate an output LRC file with line level timestamps only.

You'd only need the corrected_segments.text, corrected_segments.start_time, and corrected_segments.end_time values - see format below for example:

  "corrected_segments": [
    {
      "id": "9v1iai",
      "text": "You know that I would let you drown me",
      "words": [...],
      "start_time": 20.580482826576578,
      "end_time": 23.30054664039039
    },

I'm also very much open to PRs implementing additional output formats, if you're keen to implement line-level LRC (which should be pretty easy to do).

One of my own use cases actually needs an additional (simple, mostly plain text) output lyrics format (Power Karaoke TXT) so I'll be working on that area soonish anyway.

There's an OutputConfig object here which can be used to make new output formats configurable: https://github.com/nomadkaraoke/python-lyrics-transcriber/blob/main/lyrics_transcriber/core/config.py

Creating a new output generator by copying and modifying the PlainTextGenerator (https://github.com/nomadkaraoke/python-lyrics-transcriber/blob/main/lyrics_transcriber/output/plain_text.py) to create something like LineLevelLRCGenerator would probably be the easiest approach.

If you wanna chat about your project or pair on anything, feel free to book a slot in my calendar any time :)
https://cal.com/beveradb

-Andrew

[ben-umble]

Thanks @beveradb for the quick reply! And sorry for my slow one (real life getting in the way here!).

The ASS format isn't supported for my use case unfortunately so looks like I need a line-level LRC file (thanks for the correct terminology here!).

I'll see if I can get time to submit a PR for it, unless you happen to get to it sooner than I do.

[beveradb]

All good, sometimes I take literally months to reply 😅

Gotcha, will see if I can easily make it output a line-level LRC too, shouldn't be too difficult.

FLAC (and pretty much any other audio format you can name) is already supported as it uses pydub (wrapper around ffmpeg)