[{"data":1,"prerenderedAt":257},["ShallowReactive",2],{"project-38157524":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":29,"sentiment":30,"stance_mix":31,"scored_comments":35,"aspects":36,"strip":40,"post_text":42,"what":43,"said":71,"maker_in_thread":156,"shape":162,"grid":166,"attempts":220,"same_week":221,"maker_shows":157,"maker_hidden":249,"reader_version":250,"card":251},"38157524","LLaVaVision","An AI \"Be My Eyes\"-like web app with a llama.cpp backend","title","Show HN: LLaVaVision: An AI \"Be My Eyes\"-like web app with a llama.cpp backend","https:\u002F\u002Fgithub.com\u002Flxe\u002Fllavavision","github.com","lxe",1699232108,2023,154,19,396,3109,13,"notable","ai","vision language model app","web_app",[24,25,26,27,28],"llava","llama.cpp","multimodal","image description","vision ai","curious",-0.043,{"positive":32,"neutral":33,"mixed":34,"critical":34},1,5,2,10,{"idea":37,"performance":39},{"positive":38,"critical":38},0,{"positive":38,"critical":34},[41,41,38,41,32,38,38,38,38,32,41,38,41,38,41,41,38,33,38],3,"A simple mobile web app inspired by Fuzzy-Search\u002Frealtime-bakllava that uses llama.cpp server backend with multimodal mode to describe and narrate what the phone camera sees.\n\nI built this thing in a few hours using a single ChatGPT thread to generate most things for me and iterate on this project. Here's the workflow: https:\u002F\u002Fchat.openai.com\u002Fshare\u002Fea84ec69-5617-45e8-8772-ac2dcfa2e1b1",{"problem":44,"audience":50,"pricing":50,"motivation":51,"time_to_build":55,"stack":59,"openness":70,"stage":70,"team_size":70},{"value":45,"source":46,"evidence":47},"Describes and narrates what the phone camera sees, like Be My Eyes","maker",[48],{"id":49},"post",null,{"value":52,"source":46,"evidence":53},"Inspired by realtime-bakllava; built as a quick experiment",[54],{"id":49},{"value":56,"source":46,"evidence":57},"A few hours, using a single ChatGPT thread to generate most code",[58],{"id":49},[60,64],{"value":61,"source":46,"evidence":62},"llama.cpp server backend with multimodal mode",[63],{"id":49},{"value":65,"source":46,"evidence":66},"BakLLaVA model",[67],{"id":68,"by":11,"text":69},"38159645","This is an awesome idea! Yeah refining and asking more questions about the image could make it viable. The BakLlava model is pretty impressive, but I do need to tune the prompts…","unknown",{"praise":72,"objections":93,"requests":111,"questions":125,"alternatives":140},[73,89],{"value":74,"source":75,"evidence":76},"Cool, useful idea for accessibility","commenter",[77,81,85],{"id":78,"by":79,"text":80},"38161903","anon23432343","Wait an actual cool peace tech on the HN front page? What happened here 0_0 Cool peace of tech!! keep on doing great work.",{"id":82,"by":83,"text":84},"38159572","enoch2090","This is cool! One thing that can be easily improved is that the narrator is still trying to describe everything, while people using Be My Eyes they usually got specific purposes…",{"id":86,"by":87,"text":88},"38160951","umtksa","awesome, love the idea and maybe putting a default prompt like: \"this images comes from the persons POV that ask you questions\" instead of \"a person holding something\" \"you are…",{"value":90,"source":75,"evidence":91},"Interest in whisper.cpp and task-specific prompts",[92],{"id":82,"by":83,"text":84},[94,98,104],{"value":95,"source":75,"evidence":96},"Narrator describes everything rather than focusing on user's purpose",[97],{"id":82,"by":83,"text":84},{"value":99,"source":46,"evidence":100},"Model hallucinates as descriptions go on",[101],{"id":102,"by":11,"text":103},"38158494","I noticed that it \"hallucinates\" in the most direct sense of the word as the description goes on.",{"value":105,"source":75,"evidence":106},"Name borrows from Be My Eyes",[107],{"id":108,"by":109,"text":110},"38161183","devinprater","A \"Be My Eyes web app\"? Well whatever. I guess Be My Eyes is just gonna be known for its Be My AI anyway. As soon as Llava is able to read text, this is gonna be pretty nice. I…",[112,116,121],{"value":113,"source":75,"evidence":114},"Add text reading once LLaVA can read text",[115],{"id":108,"by":109,"text":110},{"value":117,"source":75,"evidence":118},"Support query inputs and refined prompts",[119,120],{"id":82,"by":83,"text":84},{"id":86,"by":87,"text":88},{"value":122,"source":75,"evidence":123},"Use for video games and inaccessible apps",[124],{"id":108,"by":109,"text":110},[126,133],{"value":127,"source":75,"evidence":128},"How does model accuracy compare to GPT-4 Vision?",[129],{"id":130,"by":131,"text":132},"38159915","FrenchDevRemote","Do you know how the model accuracy compares to GPT-4 vision?",{"value":134,"source":75,"evidence":135},"Can it identify exact brands and models?",[136],{"id":137,"by":138,"text":139},"38171814","ReD_CoDE","Is there any solution that just identifies the object as accurate as possible? For instance when it sees a t-shirt, can be identify the exact brand and model?",[141,148,152],{"value":142,"source":75,"evidence":143},"GPT-4 Vision \u002F ChatGPT",[144,147],{"id":145,"by":11,"text":146},"38177863","Yes, GPT Vision \u002F ChatGPT is very accurate. CogVLM is also very powerful as a self-hosted solution.",{"id":130,"by":131,"text":132},{"value":149,"source":46,"evidence":150},"CogVLM",[151],{"id":145,"by":11,"text":146},{"value":153,"source":75,"evidence":154},"Be My Eyes \u002F Be My AI",[155],{"id":108,"by":109,"text":110},{"comments":157,"answered":34,"key":158},4,[159,160,161],{"id":68,"text":69},{"id":102,"text":103},{"id":145,"text":146},{"commenters":163,"max_depth":41,"maker_replies":157,"first_comment_mins":164,"lifespan_hours":165},15,69,39,[167,169,170,172,174,178,179,183,187,190,193,194,196,198,201,205,209,214,217],[108,38,41,168,109,38],"project\u002Fmixed\u002Fidea",[82,32,41,168,83,38],[68,32,38,171,11,32],"maker\u002Fpositive\u002Fidea",[86,32,41,173,87,38],"project\u002Fneutral\u002Fidea",[175,34,32,176,177,38],"38157949","project\u002Fcritical\u002Fperformance","yousnail",[102,34,38,176,11,32],[180,34,38,181,182,38],"38160077","commenter\u002Fneutral\u002Fperformance","ggerganov",[184,34,38,185,186,38],"38159171","other\u002Fneutral\u002Fnone","DANmode",[188,34,38,185,189,38],"38159898","tsunamifury",[191,34,32,176,192,38],"38160613","darylteo",[137,41,41,173,138,38],[145,41,38,195,11,32],"maker\u002Fneutral\u002Fperformance",[130,157,41,197,131,38],"project\u002Fneutral\u002Fperformance",[199,157,38,200,11,32],"38164091","maker\u002Fmixed\u002Fperformance",[202,33,41,203,204,38],"38161000","project\u002Fneutral\u002Fnone","thih9",[206,33,41,207,208,38],"38161854","maker\u002Fneutral\u002Fnone","leodriesch",[210,211,38,212,213,38],"38159481",6,"other\u002Fpositive\u002Fdocs","birdyrooster",[78,215,33,216,79,38],7,"project\u002Fpositive\u002Fnone",[218,215,38,219,79,38],"38162174","commenter\u002Fpositive\u002Fnone",[],[222,227,231,236,240,244],{"id":223,"name":224,"points":225,"category":226},"38167363","Inshellisense",388,"dev_tools",{"id":228,"name":229,"points":230,"category":226},"38161452","Ladder",377,{"id":232,"name":233,"points":234,"category":235},"38155544","WireHub",312,"infrastructure",{"id":237,"name":238,"points":239,"category":226},"38152109","Sshx",242,{"id":241,"name":242,"points":243,"category":20},"38194586","Bulletpapers",178,{"id":245,"name":246,"points":247,"category":248},"38108044","DoltgreSQL",162,"data",false,"v2",{"id":252,"name":5,"line":6,"ai":249,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":253,"top":18,"kw":254,"maker":11,"domain":10,"tier":19,"attempts":38,"strip":255,"url":9,"mix":256,"praise":74,"objection":95,"sub":21},38157524,-0.04,[24,25,26,27,28],"3303100001303033050",[32,33,34,34],1794036644339]