SentrySearch – Semantic search over dashcam footage

Name: SentrySearch – Semantic search over dashcam footage
Availability: InStock
Author: sohamrj

by sohamrj·Mar 18, 2026·2 points·0 comments

Visit Project View on HN

AI Analysis

●●●BangerBig BrainNiche Gem

Native video embedding beats captioning pipelines — Gemini 2 does the heavy lifting.

Strengths

•Direct video-to-text embedding skips the transcription bottleneck entirely.
•ChromaDB indexing with ffmpeg auto-trim makes results immediately usable.
•$2.50/hr indexing cost is reasonable for the capability you get.

Weaknesses

•Locked into Google's Gemini API — no local or alternative model support.
•Still frames and scene detection optimizations are marked as TODO.

Post Description

I built a semantic search CLI for dashcam footage using Gemini Embedding 2's native video embedding.

The interesting part: Gemini Embedding 2 projects raw mp4 video directly into the same vector space as text, no captioning or transcription pipeline. You embed 30-second video chunks as RETRIEVAL_DOCUMENT, embed a text query as RETRIEVAL_QUERY, and cosine similarity just works across modalities.

The tool splits footage into overlapping chunks, indexes them in a local ChromaDB instance, and auto-trims the top match from the original file via ffmpeg.

Cost is about $2.50/hr of footage to index, queries are negligible. Definitely room to optimize: skipping still frames, scene detection for smarter chunking, etc.