Margin - Turn the YT videos you watch into your next video idea

byβ€’
Creators face two problems: you forget what you watch, and you run out of ideas. AI idea tools give generic filler. Analytics tell you what already happened. Note apps become a graveyard you never reopen. Margin connects them. It turns every video you watch into a clean Google Doc notes, timestamps, takeaways then mines your watch history for your next ideas: hooks, titles and angles built for your audience, not generic AI. Built solo, early β€” honest feedback wanted.

Add a comment

Replies

Best
Maker
πŸ“Œ
Hey Product Hunt πŸ‘‹ I'm Chandan, solo maker of Margin. What inspired it: I kept running into the same wall as a creator not filming, not editing, but deciding what to make next. I'd watch a ton of videos in my niche, and every one sparked something: a hook, an angle, a "I'd do this completely differently." And every one of those sparks died in a notes app I never reopened. My ideas had a graveyard, and it was my Notes app. The problem I set out to solve: the blank page. Creators don't lack ideas they lose them. The good stuff shows up while you're watching, and there was no tool that captured that and turned it back into something you could actually make. How the approach evolved: I started building a note-taker β€” "turn videos into clean summaries." But summaries just became a nicer graveyard. Nobody reopens notes. The real unlock came when I flipped it: the notes aren't the product, they're the fuel. So Margin now watches what you watch, remembers what mattered, and pushes ideas back out to you when you're stuck grounded in your actual watch history, not generic AI prompts. That shift, from "storage" to "your next idea," is the whole product now. It's early and built solo, so it's rough in places. I'm here all day tell me what's confusing, what's missing, and whether the idea-generation actually feels useful or gimmicky. Brutal feedback is the best gift today. πŸ™

how does it decide which takeaways actually matter and which get ignored?

Β It's not just "grab the longest sentences." The transcript goes to the model with a job: pull the points a viewer would actually want to remember, not every fact stated. So it weights things like is this a claim/insight vs. filler, does the speaker land on it and build around it, is it a takeaway you'd act on vs. a passing aside. Timestamps come from where that point actually happens so you can jump back and judge for yourself. It's not perfect yet if you ever get a doc where it kept the fluff and dropped the gold, send me the video, that's exactly the feedback that makes it sharper. πŸ™

the auto generated doc with timestamps saves me from scribbling half notes while watching youtube, and pulling hooks from my own history actually feels like a real starting point instead of those generic ai prompts

Β  That's exactly the itch I built it for the half-scribbled notes while half-watching. πŸ˜… Curious: when you pull hooks from your history, would you want it to lean more "safe/proven" or "weird/risky" angles? Trying to figure out the right default.

The notes came out cleaner than I expected and actually picked up the timestamps I cared about, not just a transcript dump. The hook suggestions from my own watch history were way more useful than the usual AI slop.

Β "Cleaner than a transcript dump" is literally the bar I was aiming for, so that means a lot πŸ™ And good to hear the hooks beat the usual AI slop that's the part I was most nervous about. What kind of channel are you running? Curious if it read your niche right.

how does it actually grab my watch history, is it pulling from youtube directly or do i have to feed it in somehow