Show HN: Gemini can now natively embed video, so I built sub-second video search

#1

Post by sohamrj » Tue, Mar 24, 2026, 2:58 PM UTC

Gemini Embedding 2 can project raw video directly into a 768-dimensional vector space alongside text. No transcription, no frame captioning, no intermediate text. A query like "green car cutting me off" is directly comparable to a 30-second video clip at the vector level.

I used this to build a CLI that indexes hours of footage into ChromaDB, then searches it with natural language and auto-trims the matching clip. Demo video on the GitHub README. Indexing costs ~$2.50/hr of footage. Still-frame detection skips idle chunks, so security camera / sentry mode footage is much cheaper.

Show HN: Gemini can now natively embed video, so I built sub-second video search
github.com

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#2

Post by ygouzerh » Tue, Mar 24, 2026, 4:07 PM UTC

That's quite interesting, well done! I haven't thought of this use case for embeddings. It open the door to quite many potential applications!

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#3

Post by dev_tools_lab » Tue, Mar 24, 2026, 4:11 PM UTC

Nice use of native video embedding. How do you handle cases where Gemini's response confidence is low? Do you have a fallback or threshold?

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#4

Post by mdrzn » Tue, Mar 24, 2026, 4:23 PM UTC

Very interesting (not for a dashcam, but for home monitoring).

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#5

Post by klntsky » Tue, Mar 24, 2026, 4:27 PM UTC

why not skip the text conversion? is it usable at all?

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#6

Post by sohamrj » Tue, Mar 24, 2026, 4:32 PM UTC

klntsky wrote: post #5

why not skip the text conversion? is it usable at all?

gemini embedding 2 converts straight video to vectors. in this case, dashcam clips don't have audio to transcribe and even if they did, it would be useless in the search

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#7

Post by stavros » Tue, Mar 24, 2026, 4:34 PM UTC

ygouzerh wrote: post #2

That's quite interesting, well done! I haven't thought of this use case for embeddings. It open the door to quite many potential applications!

Man, the surveillance applications for this are staggering.

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#8

Post by Aeroi » Tue, Mar 24, 2026, 4:35 PM UTC

very cool, anybody have apparent use cases for this?

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#9

Post by sohamrj » Tue, Mar 24, 2026, 4:37 PM UTC

dev_tools_lab wrote: post #3

Nice use of native video embedding. How do you handle cases where Gemini's response confidence is low? Do you have a fallback or threshold?

as of now, no threshold but that is planned in the future.

for example, for now if i search "cybertruck" in my indexed dashcam footage, i don't have any cybertrucks in my footage, so it'll return a clip of the next best match which is a big truck, but not a cybertruck

Re: Show HN: Gemini can now natively embed video, so I built sub-second video search

#10

Post by sohamrj » Tue, Mar 24, 2026, 4:48 PM UTC

Aeroi wrote: post #8

very cool, anybody have apparent use cases for this?

dashcam and home security footage are the 2 main ones i can think of.

a bit expensive right now so it's not as practical at scale. but once the embedding model comes out of public preview, and we hopefully get a local equivalent, this will be a lot more practical.