4 ms·
Could you speak more on the Shazam style part? Don't you need a database of all ads to figure out if some snippet is an ad or not? That's how Shazam works for s
by satvikpendem 3mo ago
Could you speak more on the Shazam style part? Don't you need a database of all ads to figure out if some snippet is an ad or not? That's how Shazam works for songs at least.
A native implementation I had thought of was running the audio through a STT LLM that then detects the ad timestamps then returns that to the UI to block.
- hamza_q_ 3mo agoYeah that's what I do (STT -> LLM -> timestamps to block), but if you did that for every episode download, due to location-targeted dynamic ads, your LLM API bill would run up very fast. So that's where the fingerprints matching comes in, for further versions of an episode (version = same episode but with different dynamic ads inside). I explained how the full system works here (someone emailed me and asked): https://pastebin.com/raw/r2YUEkK5 https://pastebin.com/raw/r2YUEkK5