VideoStance cross-analyses claims from expert AI videos to show you where creators agree, where they disagree, and who has the unique insight β all attributed to the original source and timestamp. Works with both YouTube and Bilibili content. No more watching 10 videos to form an opinion.
No reviews yetBe the first to leave a review for VideoStance
Maker
π
Building VideoStance started with a simple frustration β when I was researching "which AI coding tool should I use," I'd watch 5-10 YouTube reviews and walk away with completely conflicting opinions. One creator loved Cursor, another swore by Claude Code. I couldn't find a tool that actually cross-referenced these takes to show me who was right.
So I built one. VideoStance extracts claims from knowledge videos and cross-references them across creators to show consensus, controversy, and unique insights β all attributed back to the original video with timestamps.
Currently covering AI coding tools, LLM comparisons, and tech investing topics, powered by an automated pipeline that processes both YouTube and Bilibili content.
Would love to hear how you currently research tools β do you watch multiple reviews and try to piece things together yourself?
Report
How does it actually handle creators who cover the same topic but use totally different terminology or framing, does it match claims semantically or just look for keyword overlap?
Report
Maker
@tunahankutlubayΒ We match claims semantically, not by keyword. The underlying model understands that "response time is unusable for real-time work" and "latency is terrible" are saying the same thing, so they get grouped into the same analysis even when the wording is completely different.
The reality is it's not perfect β sarcasm and metaphor can still trip it up. But using the full transcript with adequate context handles the terminology drift pretty well in practice.
Report
Saved me an hour of scrubbing through three different creator videos just to compare their takes on a topic. The timestamp links back to the exact moment in each video, which is genuinely clever for fact-checking.
Report
Maker
@sedefm1cbΒ That's exactly why we built it β the timestamp-to-source thing was one of those "why doesn't this exist already" moments. Glad it's saving you real time.
Report
Pulled it up on a few AI policy clips and the timestamp links back to exact quotes, which saved me a ton of scrubbing. Genuinely useful for settling debates without sitting through every full video.
Report
Maker
@berfindang69225Β "settling debates without sitting through every full video" β that's a great way to put it, might steal that. Appreciate the kind words.
Report
How does it actually determine when two creators are talking about the same claim, especially when they use totally different phrasing or jump around topics?
How does it actually handle creators who cover the same topic but use totally different terminology or framing, does it match claims semantically or just look for keyword overlap?
@tunahankutlubayΒ We match claims semantically, not by keyword. The underlying model understands that "response time is unusable for real-time work" and "latency is terrible" are saying the same thing, so they get grouped into the same analysis even when the wording is completely different.
The reality is it's not perfect β sarcasm and metaphor can still trip it up. But using the full transcript with adequate context handles the terminology drift pretty well in practice.
Saved me an hour of scrubbing through three different creator videos just to compare their takes on a topic. The timestamp links back to the exact moment in each video, which is genuinely clever for fact-checking.
@sedefm1cbΒ That's exactly why we built it β the timestamp-to-source thing was one of those "why doesn't this exist already" moments. Glad it's saving you real time.
Pulled it up on a few AI policy clips and the timestamp links back to exact quotes, which saved me a ton of scrubbing. Genuinely useful for settling debates without sitting through every full video.
@berfindang69225Β "settling debates without sitting through every full video" β that's a great way to put it, might steal that. Appreciate the kind words.
How does it actually determine when two creators are talking about the same claim, especially when they use totally different phrasing or jump around topics?