What are these weird transcriptions?

I mean... Yesterday I posted an instrumental, completely lyric-free piece of music in a topic, and apparently the speech recognition of Elten keeps saying in Russian, "Welcome to Kazakhstan." What model does Elten’s transcription system use? Because I have absolutely nothing against Kazakhstan, but there wasn't a single word spoken in that song. :D

I don't know, but I post some certain audio files, but the transcription always says, thanks for watching, over and over and over again

It is Whisper from OpenAI. And, yes, it behaves in rather... interesting way with music and ambients. :)

Interesting?

I mean hey at leasts it's better than youtube's heat heat BS

I think it's implying that the elten community should meet up in astana. Anyone else down?

not a bad idea
-- (starchild):
I think it's implying that the elten community should meet up in astana. Anyone else down?
--

Hmmm. Does their API not have voice activity detection? Any sane whisper frontend that you can find for desktop has one of those.