[{"data":1,"prerenderedAt":566},["ShallowReactive",2],{"project-47503617":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":28,"sentiment":29,"stance_mix":30,"scored_comments":35,"aspects":36,"strip":44,"post_text":48,"what":49,"said":78,"maker_in_thread":212,"shape":221,"grid":226,"attempts":523,"same_week":531,"maker_shows":33,"maker_hidden":558,"reader_version":559,"card":560},"47503617","sentrysearch","Gemini can now natively embed video, so I built sub-second video search","title","Show HN: Gemini can now natively embed video, so I built sub-second video search","https:\u002F\u002Fgithub.com\u002Fssrajadh\u002Fsentrysearch","github.com","sohamrj",1774364307,2026,438,108,67,4366,2,"legend","ai","video search","cli",[21,24,25,26,27],"semantic search","gemini embedding","dashcam footage","vector search","loved",0.288,{"positive":31,"neutral":32,"mixed":33,"critical":34},15,35,1,0,51,{"idea":37,"pricing":39,"docs":40,"performance":41,"execution":42,"privacy":43},{"positive":38,"critical":34},10,{"positive":34,"critical":34},{"positive":33,"critical":34},{"positive":33,"critical":34},{"positive":34,"critical":34},{"positive":34,"critical":34},[45,45,46,34,34,45,45,34,34,46,34,46,46,46,46,45,45,45,45,47,46,46,45,45,45,46,45,46,45,46,45,45,45,45,34,34,34,34,34,46],3,5,4,"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.\n\nI 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.\nIndexing costs ~$2.50\u002Fhr of footage. Still-frame detection skips idle chunks, so security camera \u002F sentry mode footage is much cheaper.",{"problem":50,"audience":56,"pricing":57,"motivation":56,"time_to_build":56,"stack":61,"openness":76,"stage":77,"team_size":76},{"value":51,"source":52,"evidence":53},"Searching hours of video footage by natural language description","maker",[54],{"id":55},"post",null,{"value":58,"source":52,"evidence":59},"Indexing costs about $2.50 per hour of footage via Gemini API",[60],{"id":55},[62,66,70],{"value":63,"source":52,"evidence":64},"Gemini Embedding 2",[65],{"id":55},{"value":67,"source":52,"evidence":68},"ChromaDB",[69],{"id":55},{"value":71,"source":52,"evidence":72},"ffmpeg",[73],{"id":74,"by":11,"text":75},"47512061","you're right that the code uses ffmpeg to downsample the chunks to 5fps before sending them, but that's only a local\u002Fbandwidth optimization, not what the api actually processes.…","unknown","prototype",{"praise":79,"objections":107,"requests":145,"questions":171,"alternatives":190},[80,96],{"value":81,"source":82,"evidence":83},"Native video embedding is impressive and useful for dashcam and home security footage","commenter",[84,88,92],{"id":85,"by":86,"text":87},"47505320","Aeroi","very cool, anybody have apparent use cases for this?",{"id":89,"by":90,"text":91},"47505089","mdrzn","Very interesting (not for a dashcam, but for home monitoring).",{"id":93,"by":94,"text":95},"47524780","emmitska","The cost structure here is what stands out to me. $2.50\u002Fhr makes this viable for personal or small-business use cases that would have been unthinkable a year ago — security…",{"value":97,"source":82,"evidence":98},"Natural language search of footage is a practical use case",[99,103],{"id":100,"by":101,"text":102},"47508125","rigrassm","I picked up a Rexing dash cam a few months back and after getting frustrated with how clunky it is to get footage of it, I decided to look into building something out myself to…",{"id":104,"by":105,"text":106},"47522669","QubridAI","This is one of those “oh, that’s actually a real product now” demos way more interesting than yet another chat wrapper.",[108,123,133],{"value":109,"source":82,"evidence":110},"Surveillance and privacy concerns about AI indexing camera feeds",[111,115,119],{"id":112,"by":113,"text":114},"47506611","macNchz","This is a really cool implementation—embeddings still often feel like magic to me. That said, this exact use case is sort of also my biggest point of concern with where AI takes…",{"id":116,"by":117,"text":118},"47507598","citruscomputing","It's being built as we speak. I attended at a city council meeting yesterday, discussing approving a contract for ALPR cameras. I learned about a product from the camera vendor…",{"id":120,"by":121,"text":122},"47505302","stavros","Man, the surveillance applications for this are staggering.",{"value":124,"source":82,"evidence":125},"Cost of indexing is high for continuous use",[126,130],{"id":127,"by":128,"text":129},"47513373","febed","This seems like something that would be very expensive to run. Do you have some representative figures at a particular resolution and frame rate?",{"id":131,"by":11,"text":132},"47507132","Totally valid concern. Right now the cost ($2.50\u002Fhr) and latency make continuous real-time indexing impractical, but that won't always be the case. This is one of the reasons I'd…",{"value":134,"source":82,"evidence":135},"Cost per hour claim disputed; frame rate math questioned",[136,140,144],{"id":137,"by":138,"text":139},"47510419","jimmySixDOF","How are you getting to $2.50\u002Fhr ? The price sheet says its 0.00079 per frame. https:\u002F\u002Fai.google.dev\u002Fgemini-api\u002Fdocs\u002Fpricing#gemini-embedding-2",{"id":141,"by":142,"text":143},"47511227","jjwiseman","From what I see the code downsamples video to 5 fps, so 1 hour of video is 3600 seconds * 5 fps = 18,000 frames. 18,000 frames * $0.00079\u002Fframe = $14.22. A couple dollars more…",{"id":74,"by":11,"text":75},[146,161],{"value":147,"source":82,"evidence":148},"Local or open-weight video embedding model to avoid uploading footage",[149,153,157],{"id":150,"by":151,"text":152},"47512141","rao-v","Is there a decent open video embedding model out there? I’d love to play with this without uploading video.",{"id":154,"by":155,"text":156},"47506066","kamranjon","Does anyone know of an open weights models that can embed video? Would love to experiment locally with this.",{"id":158,"by":159,"text":160},"47521001","ideashower","Is there a local model that this would work with?",{"value":162,"source":82,"evidence":163},"Threshold for low-confidence matches",[164,168],{"id":165,"by":166,"text":167},"47504887","dev_tools_lab","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?",{"id":169,"by":166,"text":170},"47515177","Makes sense for now. Thresholding becomes critical at scale though — good luck with the next iteration!",[172,179],{"value":173,"source":82,"evidence":174},"Can it search for a specific person from a photo",[175],{"id":176,"by":177,"text":178},"47514760","subhashp","Can I give it a photo of a person and ask it to search for the person in the video?",{"value":180,"source":82,"evidence":181},"Does it work on dance videos or audio",[182,186],{"id":183,"by":184,"text":185},"47512290","crashabr","I wonder how well this would work with dance videos.",{"id":187,"by":188,"text":189},"47506174","SpaceManNabs","> No transcription, no frame captioning, no intermediate text. If there is text on the video (like a caption or wtv), will the embedding capture that? Never thought about this…",[191,201,208],{"value":192,"source":82,"evidence":193},"sentinelsearch local proof of concept using Qwen3-VL embedding",[194,198],{"id":195,"by":196,"text":197},"47531215","jakejmnz","Very cool stuff, gave me the inspiration to try it locally. Works fairly well I think: https:\u002F\u002Fgithub.com\u002Fjakejimenez\u002Fsentinelsearch",{"id":199,"by":196,"text":200},"47520744","more or less works similarly, made a proof of concept for it: https:\u002F\u002Fgithub.com\u002Fjakejimenez\u002Fsentinelsearch",{"value":202,"source":82,"evidence":203},"Intel OpenVINO AI video retrieval",[204],{"id":205,"by":206,"text":207},"47513930","martz","this can be done locally https:\u002F\u002Fgithub.com\u002Fintel\u002Fopenvino-ai-video-retrieval-analysis",{"value":209,"source":82,"evidence":210},"Fusus and ALPR camera dashboards",[211],{"id":116,"by":117,"text":118},{"comments":213,"answered":214,"key":215},13,9,[216,217,218],{"id":131,"text":132},{"id":74,"text":75},{"id":219,"text":220},"47507205","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.…",{"commenters":222,"max_depth":223,"maker_replies":213,"first_comment_mins":224,"lifespan_hours":225},81,6,69,52,[227,229,231,235,239,241,242,244,247,249,252,255,258,262,265,269,272,274,277,279,282,285,288,290,293,295,298,301,304,307,310,313,314,317,320,323,326,330,331,333,335,337,340,343,344,348,353,357,361,364,367,370,374,378,380,383,385,388,392,393,395,397,398,402,404,406,409,411,412,414,418,420,422,426,428,432,436,438,440,442,444,446,449,452,455,458,461,464,466,470,472,476,478,481,485,488,491,494,497,499,502,505,508,511,514,516,519,520],[112,34,45,228,113,34],"project\u002Fmixed\u002Fidea",[131,34,34,230,11,33],"maker\u002Fneutral\u002Fpricing",[232,34,34,233,234,34],"47507588","commenter\u002Fneutral\u002Fpricing","mpalmer",[236,34,34,237,238,34],"47507635","commenter\u002Fneutral\u002Fnone","wholinator2",[137,34,45,240,138,34],"project\u002Fneutral\u002Fpricing",[141,34,45,240,142,34],[74,34,34,243,11,33],"maker\u002Fneutral\u002Fdocs",[245,34,46,246,142,34],"47513290","maker\u002Fpositive\u002Fdocs",[116,34,34,248,117,34],"other\u002Fneutral\u002Fnone",[250,34,34,248,251,34],"47517147","robertlagrant",[253,34,34,248,254,34],"47507715","cake_robot",[256,34,34,248,257,34],"47511238","whattheheckheck",[259,34,34,260,261,34],"47507807","other\u002Fneutral\u002Fprivacy","Ajedi32",[263,34,34,260,264,34],"47508182","Spivak",[266,34,45,267,268,34],"47508005","project\u002Fneutral\u002Fidea","janalsncm",[270,34,45,267,271,34],"47508251","hypeatei",[273,34,45,267,113,34],"47510328",[275,34,34,248,276,34],"47508933","greggsy",[278,34,34,248,113,34],"47510388",[280,34,34,248,281,34],"47511318","FuckButtons",[283,34,34,248,284,34],"47515123","mbokinala",[286,34,34,260,287,34],"47511687","zahlman",[289,34,34,237,113,34],"47512423",[291,34,34,260,292,34],"47512656","cjbgkagh",[100,33,46,294,101,34],"project\u002Fpositive\u002Fidea",[296,18,34,248,297,34],"47513088","npilk",[299,18,34,248,300,34],"47514224","sbinnee",[302,18,34,248,303,34],"47517366","WarmWash",[305,18,34,248,306,34],"47517975","greesil",[308,18,34,248,309,34],"47520044","CamperBob2",[311,45,45,267,312,34],"47507935","cloogshicer",[89,47,46,294,90,34],[315,47,46,294,316,34],"47511751","fhe",[318,47,45,267,319,34],"47513650","SoftTalker",[321,46,34,248,322,34],"47510430","lwarfield",[324,223,46,294,325,34],"47506848","danbrooks",[327,328,46,294,329,34],"47506518",7,"simonreiff",[219,328,34,230,11,33],[332,328,45,267,309,34],"47513178",[195,328,46,334,196,34],"project\u002Fpositive\u002Fperformance",[187,336,45,267,188,34],8,[338,336,34,339,11,33],"47506303","maker\u002Fneutral\u002Fidea",[341,214,34,248,342,34],"47505771","7777777phil",[93,38,45,240,94,34],[345,346,34,248,347,34],"47508952",11,"bobafett-9902",[349,350,45,351,352,34],"47514813",12,"project\u002Fneutral\u002Fexecution","bob1029",[354,213,45,355,356,34],"47532895","project\u002Fneutral\u002Fperformance","shivang2607",[358,359,34,248,360,34],"47514530",14,"novoreorx",[362,359,34,363,11,33],"47517675","maker\u002Fneutral\u002Fnone",[365,31,45,351,366,34],"47509946","WatchDog",[368,31,34,369,11,33],"47510466","maker\u002Fneutral\u002Fexecution",[371,372,46,294,373,34],"47506364",16,"nullbyte",[375,372,45,376,377,34],"47506445","project\u002Fneutral\u002Fprivacy","apwheele",[379,372,45,376,196,34],"47531233",[381,372,46,382,196,34],"47531226","project\u002Fpositive\u002Fnone",[127,384,45,240,128,34],17,[386,384,45,240,387,34],"47520027","addandsubtract",[389,390,46,294,391,34],"47504790",18,"ygouzerh",[120,390,34,260,121,34],[165,394,45,355,166,34],19,[396,394,34,355,11,33],"47505369",[169,394,45,355,166,34],[399,400,45,363,401,34],"47534287",20,"wuliwong",[150,403,45,376,151,34],21,[154,405,45,376,155,34],22,[407,405,34,408,11,33],"47506358","project\u002Fneutral\u002Fnone",[410,405,45,408,309,34],"47513187",[199,405,34,248,196,34],[413,405,34,248,309,34],"47526656",[415,416,46,382,417,34],"47510942",23,"cat-turner",[176,419,45,267,177,34],24,[205,421,34,248,206,34],25,[423,424,45,355,425,34],"47508144",26,"totisjosema",[427,424,34,355,11,33],"47508195",[429,430,46,382,431,34],"47528141",27,"giversoftheint",[433,434,45,363,435,34],"47512318",28,"sans_souse",[437,434,34,363,11,33],"47512341",[158,439,45,376,159,34],29,[183,441,45,267,184,34],30,[85,443,46,294,86,34],31,[445,443,34,240,11,33],"47505572",[447,443,45,267,448,34],"47505784","hebelehubele",[450,443,45,267,451,34],"47506213","wahnfrieden",[453,443,45,267,454,34],"47506029","giozaarour",[456,443,45,267,457,34],"47506227","vidarh",[459,443,34,248,460,34],"47510806","mannyv",[462,443,34,248,463,34],"47516306","iso1631",[465,443,45,267,309,34],"47513124",[467,468,45,267,469,34],"47510087",32,"thegabriele",[471,468,34,267,11,33],"47510383",[473,474,45,267,475,34],"47505177",33,"klntsky",[477,474,34,267,11,33],"47505271",[479,474,34,248,480,34],"47505729","password4321",[482,483,34,248,484,34],"47505718",34,"emsign",[486,483,34,248,487,34],"47505964","BrokenCogs",[489,483,34,248,490,34],"47506191","throwup238",[492,483,34,248,493,34],"47506362","ting0",[495,483,34,248,496,34],"47506417","RobotToaster",[498,483,34,248,306,34],"47517986",[500,483,34,248,501,34],"47506644","jama211",[503,483,34,248,504,34],"47507040","zwirbl",[506,483,34,248,507,34],"47507377","52-6F-62",[509,483,34,248,510,34],"47506910","nclin_",[512,483,34,248,513,34],"47510261","anxoo",[515,483,34,248,300,34],"47514241",[517,483,34,248,518,34],"47514550","moomoo11",[104,32,46,294,105,34],[521,522,46,294,105,34],"47509272",36,[524,527,530],{"id":525,"t":526,"points":18,"comments":34},"47427193",1773848557,{"id":528,"t":529,"points":223,"comments":34},"47440988",1773933570,{"id":4,"t":12,"points":14,"comments":15},[532,537,542,546,550,554],{"id":533,"name":534,"points":535,"category":536},"47506713","Video.js v10",648,"dev_tools",{"id":538,"name":539,"points":540,"category":541},"47505144","Email.md",380,"productivity",{"id":543,"name":544,"points":545,"category":536},"47491466","Cq",225,{"id":547,"name":548,"points":549,"category":536},"47452515","Sonar",204,{"id":551,"name":552,"points":553,"category":536},"47499672","ProofShot",161,{"id":555,"name":556,"points":557,"category":20},"47524704","cog",157,false,"v2",{"id":561,"name":5,"line":6,"ai":558,"cat":20,"year":13,"points":14,"comments":15,"rec":28,"s":562,"top":18,"kw":563,"maker":11,"domain":10,"tier":19,"attempts":45,"strip":564,"url":9,"mix":565,"praise":81,"objection":109,"sub":21},47503617,0.29,[21,24,25,26,27],"3350033005055553333455333535353333000005",[31,32,33,34],1793991149000]