I hate that everything interesting is in the videos and podcasts, so I built a little stack that downloads them with the transcripts, transcribe those with transcripts unavailable, funnels the media files to Plex library, makes transcripts searchable with Meili and pushes them into KB in OpenWebUI, so they can be queried with local LLMs. And you know what? It’s working… 👀👀👀
Replies (9)
Nice, what do you use for the transcripts? Whisper? What about pushing summaries to Nostr?
Yes, whisper. I suppose I can do summaries, but what for? Would that be useful for anyone?
I think I'm missing some good stuff, so having an episode summary ideally with times would be great. A nice way of discovering new content.
Hmmm, okay, let me add this to the backlog…
My current problem now is lack of serious firepower. Hey, nostr:npub1sceswwu9vldf0tg3nlkajutp2j4zsv35ewma9r23x8d2qa73eaaqmmq7fu, what about that 5k AI machine, I have some stuff to send your way 😂
Exactly. Running anything on my junk server is impossible or impossibly slow, and running it on my desktop is impractical. I use OpenRouter for LLM access, which is convenient, fast, and practical, but also expensive. A 5k AI machine sounds like something from a fairy tale.
If we can get 20 people paying 20 usd per month, it’s paid in a year… I’m saying, let’s do it 🤷♀️
Oh, $5k, makes sense. That does not sound like a dream anymore.
Normally I would buy a car for that money, but I already got five and I don’t have any parking space left…
