SpokenStreamSleuth

Your go-to solution for effortlessly detecting, analyzing, and labelling spoken audio streams within your Plex media library.

The Problem

Unlabelled media. At a minimum it frustrates the completionist, and at it's worst it is inaccessible and confusing to those who rely on the comfort of entertainment as an escape. If we're looking to curate a specific composition it's difficult to asses your progress without langauge tags on the audio streams contained therein.

The Solution

Retrieve a list of unknown audio streams and query Whisper for their language. Systematically update the metadata of the original file and then notify Plex that an analysis is needed.

Features

• Automated language identification
• Fixes the files that are supported, skips those that aren't
• Edits the source file itself so the solution is evident no matter what is accessing the video

Installation

If I can do it, so can you!

Prerequisites

• Python3.x
• FFmpeg
• mkvtoolnix
• ahmetoner/whisper-asr-webservice

Steps (Using a tteck Ubuntu/Debian installion as an example)

1. Install prerequisites

sudo apt update && sudo apt install git ffmpeg mkvtoolnix pip

2. Clone the repository and enter it

git clone https://github.com/MrSaiclops/SpokenStreamSleuth && cd SpokenStreamSleuth

3. Install Python requirements

pip install -r requirements.txt

4. Run Whisper webservice

docker run -d -p 9000:9000 -e ASR_MODEL=medium -e ASR_ENGINE=faster_whisper onerahmet/openai-whisper-asr-webservice:latest

5. Set config settings

nano config.txt

Instructions for getting the token here. If whisper is hosted on the same machine you should be fine to use localhost:9000

plexURL=http://192.168.1.187:32400
token=XWpR93-KLfQv5-GHsTb6
library=TV Shows
whisper=192.168.1.122:9000