No transcript of the OBS video recording output

[ltolstoy]

I use OBS v32.0.1 64bit, with local-vocal v0.4.1 with obs-localvocal-0.4.1-windows-x64-cuda. My goal is to get a separate transcript for each OBS recording file. I set up transcription_filterAudioFilter for DesctopAudio (no Mic/Aux) in Audio Mixer, use Wisper Tiny English LLM model, no translation, but don't get any txt file at output folder. OBS logs also don't have anything related to local-vocal saving the output txt transcript file. What am I missing? Please suggest.
I attach the OBS log and local-vocal settings screenshot.

2025-10-23 15-20-29.txt

[Tabby]

Thanks for the detailed report ^_^

So, there's a few things going on here. First one is that because you have the "rename_file_to_match_recording" option checked, it's renaming the SRT file to match the video name and moving it alongside the video, which might be why you can't find the SRT file.

For example, I configured the plugin in my OBS to put the SRT file in /home/tabby/obs_srt_files/test-output.srt and configured the video recordings to go in /home/tabby/Videos/obs_recordings

While recording is in progress:

After stopping the recording:

There is another issue, in that it seems that the truncate option truncates the file after every sentence not just at the start of a recording. You should probably turn that off or you'll only get the very last line in the SRT file (and I need to look into that behaviour because it seems... not right)

[ltolstoy]

Thank you Tabby for the response. I set in transcription_filterAudioFilter settings: subtitle_sources: LocalVocal Subtitles, n file_output_group: save_srt, only_whyle_recording ON, and truncate_output_file, rename_file_to_match_recording OFF; log_group: INFO. My problem is that i don't see ANY srt or txt files in any of the folders (neither in where videos are stored, nor where i expect to see srt or txt transcripts after the OBS recording session is complete). Also in OBS logs I see that local-vocal plugin is activated, but don't see any signs of it transcribed anything, or saved anything on hard drive (not sure if it is supposed to save any kind of this info in the log, but I see that mkv file is saved by ffmpeg, like [ffmpeg muxer: 'simple_file_output'] Writing file 'D:/video/video_rec/interviews and tests/2025-10-24 14-10-37.mkv').
I've tried to play with transcript name, make it fixed or dynamic like transcript_%Y-%m-%d_%H-%M-%S.txt , or turn on " rename_file_to_match_recording" - with no effect. I think I miss something simple, but don't know what. So I don't know which part is not working , and how to fix it. Any ideas?
My laptop has NVIDIA GeForce GTX 1060 with Max-Q videocard, it usually works with CUDA. May be plugin is not compatible with it and I need just CPU version?

2025-10-24 14-10-31.txt

[Tabby]

I still think you should try disabling the truncate option because that probably doesn't do what you want

But it looks like you might also have been being affected by the same issue as in #246 which should be fixed in v0.4.2 so could you please try updating the plug-in to that version and trying again?

[ltolstoy]

Thank you Tabby, the new v 0.4.2 works and outputs something, I set "Rename File to match recording" to ON, and as you said before, during the recording session there is a temporary.srt file in folder i created to save transcripts, which disappears from there after recording is complete. And then there is a srt file named the same as recorded mkv video in the folder I created to save the video.
There is a confusing part for me: If i set local-vocal plugin on "Desktop Audio" in Audio Mixer, then only external sound is transcribed (i.e sound from the youtube video or meeting, but not my own voice). If I put this plugin on "Mic/Aux", then both my voice and external sounds are transcribed. If I set up this plugin in both "Desktop Audio" and "Mic/Aux", and both have "Rename File to match recording" set to ON, it seems that only one final version of srt file is saved with mkv file in video output folder, and it seems that it is from Mic /Aux channel (or, may be it is the latest set file, now I am not sure). Is it expected behavior?
Also it seems that transcript files can have only static names like tmp.txt or tmp.srt, and names with date and time are not supported, like tmp_$T.txt (to add timestamp), or tmp_%Y-%m-%d_%H-%M-%S.txt or tmp_Y_M_D-h-m-s.txt. Is there a way to save transcript for video recording in separate folder with dynamically-formed name (with timestamp in general case)? Because for static transcript name the file stored in transcript folder will be overwritten for every new recording, right?

[Tabby]

There is a confusing part for me: If i set local-vocal plugin on "Desktop Audio" in Audio Mixer, then only external sound is transcribed (i.e sound from the youtube video or meeting, but not my own voice). If I put this plugin on "Mic/Aux", then both my voice and external sounds are transcribed. If I set up this plugin in both "Desktop Audio" and "Mic/Aux", and both have "Rename File to match recording" set to ON, it seems that only one final version of srt file is saved with mkv file in video output folder, and it seems that it is from Mic /Aux channel (or, may be it is the latest set file, now I am not sure). Is it expected behavior?

This is expected, at least at the moment. The plugin assumes it's only being used on a single audio source, and that's one of the things I'd like to improve at some point, though for the moment I'm trying to fix stability issues and getting acceleration working on all platforms before I try and improve usability or add new features

Also it seems that transcript files can have only static names like tmp.txt or tmp.srt, and names with date and time are not supported, like tmp_$T.txt (to add timestamp), or tmp_%Y-%m-%d_%H-%M-%S.txt or tmp_Y_M_D-h-m-s.txt. Is there a way to save transcript for video recording in separate folder with dynamically-formed name (with timestamp in general case)? Because for static transcript name the file stored in transcript folder will be overwritten for every new recording, right?

At the moment, no. The only way to have dynamic names is to have them renamed to match the recording, which does at least avoid the issue of overwriting the transcript. It would be good to change that config to take a folder, a name, and an extension and just always include a timestamp or something in the filename, but again that's not currently possible I'm afraid