sohamrj

8 ms·

1.
by sohamrj 3mo ago
Does this work reliably? Don't ChatGPT, Gemini, etc. change up their DOMs frequently and use bot detection to prevent this exact use case?
2.
by sohamrj 6mo ago
Last week, I posted SentrySearch, a CLI for semantic video search using Gemini's embedding API. The #1 request was local model support. Turns out Qwen3-VL-Embedding can natively embed video into the same kind of vector space, no API, f
3.
Show HN: Local video search with Qwen3-VL: no API, runs on Apple Silicon, GPUs(github.com)
3 points by sohamrj 6mo ago|0 comments
4.
by sohamrj 6mo ago
ur not wrong
5.
by sohamrj 6mo ago
haha no i'm driving the tesla and that clip is from the left repeater camera (teslas record from all around the car)
6.
by sohamrj 6mo ago
you're right that the code uses ffmpeg to downsample the chunks to 5fps before sending them, but that's only a local/bandwidth optimization, not what the api actually processes. regardless of the file's frame rate, the g
7.
by sohamrj 6mo ago
yea, it's so events on a chunk boundary still get captured in at least one chunk. i haven't had the chance to do formal benchmarks on overlap vs. no-overlap yet. the 5s default is a pragmatic choice, long enough to catch most even
8.
by sohamrj 6mo ago
dashcam is just one of the use cases and the one i tested on. but this could theoretically work with any kind of video footage like home security footage
9.
by sohamrj 6mo ago
Yeah, this is a great idea, I’ve actually been thinking about exactly this as the next logical step. SentrySearch already returns precise in/out timestamps for any natural-language query and uses ffmpeg to auto-trim clips. Turning that
10.
by sohamrj 6mo ago
I've found I have to be very specific to get the clip I'm searching for. For example, "car cuts me off" just returned a clip of a car driving past my blindspot. But, "car with bike rack on back cuts me off at night&
11.
by sohamrj 6mo ago
Thanks! Yeah that would be pretty cool, but continuous indexing would be pretty expensive now, because the model's in public preview and there are no local alternatives afaik. This very well might be a reality in a couple years though!
12.
by sohamrj 6mo ago
Totally valid concern. Right now the cost ($2.50/hr) and latency make continuous real-time indexing impractical, but that won't always be the case. This is one of the reasons I'd want to see open-weight local models for this,
13.
by sohamrj 6mo ago
Not aware of any that do native video-to-vector embedding the way Gemini Embedding 2 does. There are CLIP-based models (like VideoCLIP) that embed frames individually, but they don't process temporal video. you'd need to average f
14.
by sohamrj 6mo ago
Yes to both. The embedding is over raw video frames, so anything visible (text, signs, captions) gets captured in the vector. And Gemini Embedding 2 extracts the audio track and embeds it alongside the visual frames. So a query like 's
15.
by sohamrj 6mo ago
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
16.
by sohamrj 6mo ago
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
17.
by sohamrj 6mo ago
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
18.
Show HN: Gemini can now natively embed video, so I built sub-second video search(github.com)
438 points by sohamrj 6mo ago|108 comments
19.
Show HN: Search dashcam footage by describing what happened(github.com)
6 points by sohamrj 6mo ago|0 comments
20.
Show HN: SentrySearch – Semantic search over dashcam footage(github.com)
2 points by sohamrj 6mo ago|0 comments