Universal Summarizer Couldn't Summarize Podcast MP3

iamjameswalters

I have a direct link to a podcast taken from the XML feed:

https://afp-597195-injected.calisto.simplecastaudio.com/695767b0-cd40-4e6c-ac8c-ac6bc0df77ee/episodes/6e13ef8a-244f-41e8-bf58-dfd0c94e1798/audio/128/default.mp3

But when I try to summarize this, I get "We are sorry, we are unable to extract the source.

Perhaps the simplecastaudio server is validating user agents or something? I dunno.

Thanks for taking a look!

leftium

I don't think the summarizer can process non-text files.

It works for YouTube videos because it downloads the transcripts and summarizes those.

So you need to submit a text version (transcript) of the podcast, not the mp3 audio.

Perhaps a more informative error message would help. (And "summarize anything in a jiffy!" does make it sound like mp3 audio should work.)

iamjameswalters

It does say in the docs this is possible.

And I have done it before with other things--I think they've got a transcription step somewhere in the pipeline (or they feed it to a multi-modal model). I think there must be something weird about simplecastaudio's server or how Kagi requests it.

Thibaultmol

I'm wondering if the transcript pipeline is broken.
https://kagi.com/summarizer?target_language=&summary=takeaway&url=https%3A%2F%2Fcommons.wikimedia.org%2Fwiki%2FFile%3ACross_of_Gold_speech_excerpt.mp3
also doesn't work

quinncom

I just tested with today’s Kagi update, and this works now.