On-the-fly transcription... possible-- worth it?
I'm trying to solve a very specific Plex annoyance and wanted to see if anyone in the community has built a script or workflow for this.
We’ve all watched a movie that is 95% English, but has a 2-minute scene where characters speak a foreign language. If you don't have the "Forced" subtitle track installed, you're stuck pausing the movie, digging through sub menus, or guessing what they said.
Here is my ideal dream setup/concept:
Some sort of background utility or webhook listener that monitors my active Plex session. It sees what movie I'm playing, looks at my current timestamp, and reads ahead roughly 2 to 5 minutes into the audio stream. If its audio analysis detects non-English dialogue coming up, it hands that specific time block to a local AI engine (like OpenAI Whisper) to auto-transcribe/translate it into a localized .forced.en.srt file and injects it. By the time my playback reaches that scene, the subtitle is ready and waiting.
I know the standard answer is "Just automate Bazarr to grab forced tracks" or "Use Subgen/Whisper to transcribe the entire movie ahead of time."
But I'm curious about the just-in-time / look-ahead aspect. Has anyone experimented with Tautulli webhooks or Plex API tracking to trigger smart, chunk-based Whisper translations on the fly while watching a movie? Or is everyone strictly pre-processing their entire library?
Curious to hear how you guys handle the "missing forced subtitles" problem using local AI!