This repository contains scripts used for my BossmanJack live stream capturing system and associated post processing.

Scripts

• kick_capture/main.py listens for the "Start Stream" event from Kick and runs the download script
• kick_capture/bmj_dl.bat runs yt-dlp and auto-restarts it in the event of errors
• kick_capture/Test-FfmpegDied.ps1 is used by the batch script to check if ffmpeg encountered an error while capturing and restart the capture if necessary. Note this used to be very frequent, but it seems Kick has resolved the random 500 errors since December.
• Export-WavFiles.ps1 automates extracting wav audio from the streams for running whispercpp
• New-WhisperCppTranscription.ps1 automates the transcription process
• Export-WhisperCsvToTxt.ps1 creates nicely formatted txt files from the CSV files whispercpp generates
• Write-CleanYtDlpMetadata.ps1 cleans the .info.json files generated by yt-dlp, stripping potentially sensitive information
• grab_kick_vods/grab_kick_vods.py grabs all Kick VODs for a channel and runs a companion script
• grab_kick_vods/grab_kick_vod.bat example batch script for retrieving a VOD if it hasn't already been downloaded
• rapfame_dl/rapfame_dl.py mass download tracks from Rap Fame by artist ID

Known Issues

• The PowerShell CSV cmdlets are buggy when dealing with escaped double quoted strings in CSV files. Whispercpp will occassionally emit quoted text and this can lead to slight errors in the txt files.
• PowerShell's JSON cmdlets have a shallow depth by default, which leads to some heavily nested objects in the yt-dlp metadata becoming stringified. This is easily fixable by specifying a sufficient depth, but the property affected isn't important (HTTP headers) and I don't want to change the schema of my existing metadata files.
• There is no handling of whispercpp or ffmpeg errors
• The Rap Fame script doesn't support obtaining the artist ID from the page