Hello everyone,
with the recent Netflix acquisition of the video episode feed we have lost the ad-free youtube feed. I hate listening to ads, so I have decided to build an ad blocker for Behind the bastards specifically, but i think it should work for most podcasts which follow the same structure eventually.
Currently, I can reliably split an episode into segments where each segment is either ad or content. The main issue is distinguishing ad from content. In some episodes you could sort them by average volume alone, but especially newer episodes this doesn't work. Segment length could help, but with the cold opens this is also unreliable, and i want to avoid "friendly fire" as much as possible. I am currently considering building on top of, or creating from scratch, some sort of machine-learning-based solution. If you have any ideas, they will be appreciated.
I would like to have a sample of episodes with a wide array of ads inserted, so i can fine-tune whichever sorting algorithm i settle on. Ads are inserted into the episodes based on (among other things) your location, and me being in the EU get mostly the same handful of ads for the same podcasts on repeat.
How you can help
Send me full episodes with ads. Especially if you have older episodes downloaded. It can be BTB or ICHH, live episodes are also good to have, as i suspect they may be tricky to also sort well.
If people are interested, will update you on the progress and open-source the project when it's in a working state.
6 Comments
vinceman@lemmy.blahaj.zone · 3 pts · 182d
Am I dumb or have we not been waiting like 3 or 4 years for Android ad free.
cheat700000007@lemmy.world · 2 pts · 182d
Why does sponsor block and ublock origin not work here?
flx@lemmy.blahaj.zone · 4 pts · 182d
UBO is for website content, not audio. Sponsorblock works, (although i think youtube is also working on dynamic in-content ads). What i want to do is have a script that crunches a file directly from the podcast feed, including old episodes not on youtube, and removes the embedded ads. I listen mostly on my phone using a podcast app and sponsorblock is not really an option there.
cheat700000007@lemmy.world · 1 pts · 182d
UBO does work for the ads inserted in to YouTube and Firefox Mobile supports both of the add-ons plus background play. Not exactly a fix for you if you like the podcast app but it's an option.
RIotingPacifist@lemmy.world · 1 pts · 137d
How do you want the episodes sent?
I'm super curious about this as i've thought of doing the same thing in a similar way.
Honestly if you have a way to mark segments, that's good enough for me, especially as if you're wrong I can go back and listen to the skipped segment.
I don't think you need to overthink it, the logic is pretty much.
If a segment is short it's an ad, unless it's the first segment and followed by a 2nd short segment (cold open).
And maybe there is a situation where 3 consecutive short segments at the end include 1 non-ad closing segment.
flx@lemmy.blahaj.zone · 1 pts · 114d
Hi, a link to a (google drive/temp file host site) will do fine, or email is also an option. Sorry for taking so long to respond, bit busy at the moment, but i'll release the code when i have some time to spare. I have segment marking working pretty much 100% of the time, so maybe i'll make it a standalone function to then extend by categorizing. Thanks for your interest