[{"data":1,"prerenderedAt":551},["ShallowReactive",2],{"project-40805010":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":36,"aspects":37,"strip":48,"post_text":50,"what":51,"said":97,"maker_in_thread":218,"shape":228,"grid":233,"attempts":513,"same_week":514,"maker_shows":40,"maker_hidden":544,"reader_version":545,"card":546},"40805010","Cerebrium voice agent (fastvoiceagent)","Voice bots with 500ms response times","title","Show HN: Voice bots with 500ms response times","https:\u002F\u002Ffastvoiceagent.cerebrium.ai\u002F","fastvoiceagent.cerebrium.ai","kwindla",1719438663,2024,315,99,186,3515,6,"legend","ai","voice agent","web_app",[24,25,26,27,28],"voice bot","low latency","speech to speech","llm voice","real-time voice ai","mixed",0.27,{"positive":32,"neutral":33,"mixed":34,"critical":35},25,26,1,9,61,{"performance":38,"idea":41,"docs":43,"execution":44,"design":46,"pricing":47},{"positive":39,"critical":40},12,3,{"positive":40,"critical":42},0,{"positive":34,"critical":42},{"positive":45,"critical":40},4,{"positive":34,"critical":40},{"positive":34,"critical":42},[49,42,42,42,42,42,49,49,40,34,49,34,45,40,34,49,49,42,42,49,40,49,49,42,40,49,34,45,40,45,40,40,40,49,49,40,40,49,49,40],5,"Last year when GPT-4 was released I started making lots of little voice + LLM experiments. Voice interfaces are fun; there are several interesting new problem spaces to explore.\n\nI'm convinced that voice is going to be a bigger and bigger part of how we all interact with generative AI. But one thing that's hard, today, is building voice bots that respond as quickly as humans do in conversation. A 500ms voice-to-voice response time is just barely possible with today's AI models.\n\nYou can get down to 500ms if you: host transcription, LLM inference, and voice generation all together in one place;",{"problem":52,"audience":58,"pricing":66,"motivation":67,"time_to_build":66,"stack":71,"openness":94,"stage":95,"team_size":96},{"value":53,"source":54,"evidence":55},"Voice bots respond too slowly to feel like human conversation","maker",[56],{"id":57},"post",{"value":59,"source":60,"evidence":61},"Engineers building voice bots","commenter",[62],{"id":63,"by":64,"text":65},"40810114","za_mike157","Hi Vessenes From Cerebrium here. Really appreciate the feedback - glad you had a good experience! This application is easy to extend\u002Fimplement meaning you can edit it to however…",null,{"value":68,"source":54,"evidence":69},"Make voice bots respond as quickly as humans in conversation",[70],{"id":57},[72,76,83,90],{"value":73,"source":54,"evidence":74},"Cerebrium hosting (GPU)",[75],{"id":57},{"value":77,"source":54,"evidence":78},"Opus audio codec",[79,80],{"id":57},{"id":81,"by":11,"text":82},"40807358","Yes, Opus is the fastest and best option for real-time audio. It was designed to be flexible and to encode\u002Fdecode at fairly low latencies. It sounds good for narrow-band (speech)…",{"value":84,"source":60,"evidence":85},"Silero VAD",[86],{"id":87,"by":88,"text":89},"40808193","makeitmore","Hi, I built the client UI for this and... yea, I really wanted to get Firefox working :( We needed a way to measure voice-to-voice latency from the end-user's perspective, and…",{"value":91,"source":60,"evidence":92},"Daily (WebRTC partner)",[93],{"id":63,"by":64,"text":65},"unknown","beta","company",{"praise":98,"objections":126,"requests":160,"questions":176,"alternatives":195},[99,118],{"value":100,"source":60,"evidence":101},"Very fast response, feels conversational",[102,106,110,114],{"id":103,"by":104,"text":105},"40807031","jaybrendansmith","This thing is incredible. It finished a sentence I was saying.",{"id":107,"by":108,"text":109},"40808621","isoprophlex","Jesus fuck that's fast, and I had no idea speed mattered that much. Incredible. Feels like an entirely different experience than the 5+ seconds latency with openai.",{"id":111,"by":112,"text":113},"40809089","mdbackman","Very, very impressive! It's incredibly fast, maybe too fast, but I think that's the point. What's most impressive though is how the VAD and interruptions are tuned. That was, by…",{"id":115,"by":116,"text":117},"40811597","realyashnag","This was scary fast. Neat interface and (almost) indistinguishable from a human over the phone \u002F internet. Kudos @cerebrium.ai.",{"value":119,"source":60,"evidence":120},"Interruption handling is natural",[121,122],{"id":111,"by":112,"text":113},{"id":123,"by":124,"text":125},"40808184","hackerbob","This is indeed fast! Also seems to be no issue interrupting it while speaking. Is this using WebRTC echo cancellation to avoid microphone and speaker audio mix ups?",[127,138,145,153],{"value":128,"source":60,"evidence":129},"Voice sounds robotic",[130,134],{"id":131,"by":132,"text":133},"40806795","aussieguy1234","Fast yes, but the voice sounds robotic.",{"id":135,"by":136,"text":137},"40811665","cloudking","It's literally a robot",{"value":139,"source":60,"evidence":140},"Interrupts users who pause",[141],{"id":142,"by":143,"text":144},"40810751","SubiculumCode","A chatbot that interrupts me even faster. Sorry for the sarcasm. maybe im just slow, but when I'm trying to formulate a question on the spot, I pause a lot. having the chatbot…",{"value":146,"source":60,"evidence":147},"Firefox support is flaky",[148,152],{"id":149,"by":150,"text":151},"40807379","geofffox","I use Firefox... still.",{"id":87,"by":88,"text":89},{"value":154,"source":60,"evidence":155},"Reported latency numbers do not add up",[156],{"id":157,"by":158,"text":159},"40808431","az226","Your marketing says 500 but your math says 759.",[161,169],{"value":162,"source":60,"evidence":163},"Firefox support",[164,168],{"id":165,"by":166,"text":167},"40807449","panja","I hate that everyone just develops for chromium only",{"id":87,"by":88,"text":89},{"value":170,"source":60,"evidence":171},"Open-source or self-hostable version",[172],{"id":173,"by":174,"text":175},"40815267","trueforma","I too am excited about voice inferencing. I wrote my own Websocket Faster whisper implementation before OpenAI's gpt4o release . They steamrolled my interview coach concept…",[177,184],{"value":178,"source":60,"evidence":179},"Total cost of running the system",[180],{"id":181,"by":182,"text":183},"40809312","c0brac0bra","I've been developing with Deepgram for a while, and this is one of the coolest demos I've seen with it! I am curious about total cost to run this thing, though. I assume that on…",{"value":185,"source":60,"evidence":186},"Which LLM and training data were used",[187,191],{"id":188,"by":189,"text":190},"40810835","spark_chicken","i have tried it. it is really fast! I know making a real-time voice bot is not easy with this low latency. which LLM did you use? how large LLM to make the conversation efficient?",{"id":192,"by":193,"text":194},"40825572","gsjbjt","That's awesome - can you say anything about what datasets this was trained on? I assume something specifically conversational?",[196,203,211],{"value":197,"source":60,"evidence":198},"Moshi by Kyutai",[199],{"id":200,"by":201,"text":202},"40866832","etherealG","moshi by Kyutai seems to have beaten your approach by about 500ms, and they're going to release open source. https:\u002F\u002Fwww.youtube.com\u002Flive\u002Fhm2IJSKcYvo hn discussion here:…",{"value":204,"source":60,"evidence":205},"GPT-4o voice mode",[206,210],{"id":207,"by":208,"text":209},"40808773","spuz","It's not exactly clear is this a voice-to-voice model or a voice-to-text-to-voice model? When it is finally released, OpenAI claim their GPT4o audio model will be a lot faster at…",{"id":107,"by":108,"text":109},{"value":212,"source":60,"evidence":213},"Siri",[214],{"id":215,"by":216,"text":217},"40808072","andrewstuart","Damned impressive. Apple's Siri still can't allow me to have a conversation in which we aren't tripping over each other and pausing and flunking and the whole thing degrades into…",{"comments":45,"answered":219,"key":220},2,[221,224,225],{"id":222,"text":223},"40810565","Full technical write-up here: https:\u002F\u002Fwww.daily.co\u002Fblog\u002Fthe-worlds-fastest-voice-bot\u002F",{"id":81,"text":82},{"id":226,"text":227},"40810722","You are right! Thank you. I went back and looked at actual benchmark numbers from a couple of years ago and the numbers I got were ~26ms one-way. I rounded up to 30 to be…",{"commenters":229,"max_depth":230,"maker_replies":45,"first_comment_mins":231,"lifespan_hours":232},74,7,310,165,[234,238,242,245,249,252,256,259,262,264,266,269,272,274,277,280,284,286,289,292,294,297,300,304,308,311,314,315,316,318,322,326,328,329,331,334,337,338,342,345,348,351,354,356,358,360,363,366,368,371,373,377,380,382,383,385,387,391,394,395,396,400,404,407,411,414,416,419,422,426,428,430,433,436,439,443,446,448,452,453,457,458,463,465,466,470,474,476,478,482,486,487,491,495,499,501,503,505,509],[235,42,49,236,237,42],"40808261","project\u002Fpositive\u002Fperformance","firefoxd",[239,42,42,240,241,42],"40808559","commenter\u002Fneutral\u002Fnone","lukan",[243,42,42,240,244,42],"40808848","jstanley",[246,42,42,247,248,42],"40809356","other\u002Fneutral\u002Fnone","sillysaurusx",[250,42,42,247,251,42],"40811531","big_man_ting",[253,42,42,254,255,42],"40817131","other\u002Fneutral\u002Fidea","dietr1ch",[257,42,42,247,258,42],"40854977","fennecbutt",[260,42,42,240,261,42],"40808956","999900000999",[263,42,42,240,241,42],"40809224",[265,42,42,240,261,42],"40814346",[267,42,42,240,268,42],"40868798","ytpete",[270,42,42,240,271,42],"40811176","asjir",[273,42,42,247,237,42],"40812762",[275,42,42,240,276,42],"40818002","leobg",[278,42,42,247,279,42],"40827788","bedel23",[281,34,49,282,283,42],"40809481","project\u002Fpositive\u002Fidea","vessenes",[63,34,40,285,64,42],"maker\u002Fneutral\u002Fdocs",[287,34,49,288,283,42],"40824377","maker\u002Fpositive\u002Fdocs",[290,34,40,291,182,42],"40811048","project\u002Fneutral\u002Fidea",[293,34,40,291,283,42],"40811827",[295,219,40,291,296,42],"40808710","luke-stanley",[298,219,42,254,299,42],"40812251","phkahler",[301,219,40,302,303,42],"40812441","project\u002Fneutral\u002Fnone","charlesyu108",[305,219,34,306,307,42],"40812887","project\u002Fcritical\u002Fperformance","regularfry",[309,40,49,236,310,42],"40808919","_def",[312,40,49,282,313,42],"40809683","danielbln",[111,45,49,236,112,42],[157,49,34,306,158,42],[317,49,42,240,255,42],"40808520",[319,49,40,320,321,42],"40809066","maker\u002Fneutral\u002Fperformance","vr000m",[323,49,40,324,325,42],"40809376","project\u002Fneutral\u002Fperformance","whizzter",[327,49,49,236,283,42],"40809500",[173,18,40,324,174,42],[330,230,42,247,271,42],"40811217",[149,332,34,333,150,42],8,"project\u002Fcritical\u002Fexecution",[335,332,42,240,336,42],"40807444","sa-code",[165,332,34,333,166,42],[339,332,49,340,341,42],"40807733","project\u002Fpositive\u002Fexecution","darren_",[343,332,42,240,344,42],"40808317","hawski",[346,332,49,340,347,42],"40808460","mcny",[349,332,42,247,350,42],"40808383","RockRobotRock",[352,332,42,240,353,42],"40808915","the8472",[355,332,42,240,350,42],"40817062",[357,332,42,240,353,42],"40819572",[359,332,42,240,350,42],"40827373",[361,332,42,247,362,42],"40808947","Bluestein",[364,332,42,240,365,42],"40813736","93po",[367,332,42,240,350,42],"40815299",[369,332,49,340,370,42],"40808144","4mitkumar",[87,332,40,372,88,42],"maker\u002Fneutral\u002Fexecution",[374,332,40,375,376,42],"40808511","project\u002Fneutral\u002Fexecution","stavros",[378,332,49,340,379,42],"40808419","chungus",[381,332,40,372,88,42],"40808479",[215,35,49,236,216,42],[115,384,49,236,116,42],10,[207,386,40,291,208,42],11,[388,386,42,389,390,42],"40808790","commenter\u002Fneutral\u002Fidea","pavlov",[392,386,42,240,393,42],"40809128","isaacfung",[222,386,42,285,11,34],[200,39,34,306,201,42],[397,398,49,282,399,42],"40807670",13,"dijit",[401,402,42,247,403,42],"40813538",14,"andrewmcwatters",[405,402,42,247,406,42],"40815108","Flumio",[408,409,49,236,410,42],"40807521",15,"anonzzzies",[142,412,34,413,143,42],16,"project\u002Fcritical\u002Fdesign",[415,412,34,413,143,42],"40810821",[417,412,40,418,143,42],"40812588","project\u002Fneutral\u002Fdesign",[420,412,40,421,88,42],"40830580","maker\u002Fneutral\u002Fdesign",[423,424,49,236,425,42],"40808445",17,"mmcclure",[123,427,40,375,124,42],18,[429,427,40,372,88,42],"40808344",[181,431,40,432,182,42],19,"project\u002Fneutral\u002Fpricing",[434,431,40,435,64,42],"40810142","maker\u002Fneutral\u002Fpricing",[437,431,49,438,182,42],"40811008","maker\u002Fpositive\u002Fpricing",[440,441,40,324,442,42],"40819556",20,"amluto",[444,441,42,240,445,42],"40820325","Olreich",[447,441,42,240,442,42],"40825919",[449,450,40,324,451,42],"40807087",21,"yjftsjthsd-h",[81,450,42,320,11,34],[454,455,40,320,456,42],"40809444",22,"yalok",[226,455,42,320,11,34],[459,460,49,461,462,42],"40815931",23,"project\u002Fpositive\u002Fnone","_DeadFred_",[103,464,49,236,104,42],24,[192,32,40,291,193,42],[467,33,40,468,469,42],"40833755","project\u002Fmixed\u002Fperformance","tamimio",[471,472,49,236,473,42],"40809771",27,"andruby",[131,475,34,413,132,42],28,[477,475,42,421,11,34],"40807016",[479,475,49,480,481,42],"40808501","project\u002Fpositive\u002Fdesign","bombela",[483,475,42,484,485,42],"40808823","commenter\u002Fpositive\u002Fnone","lofties",[135,475,40,418,136,42],[488,489,40,291,490,42],"40810907",29,"p_frank",[492,493,49,461,494,42],"40809094",30,"sumedh",[496,497,49,236,498,42],"40809257",31,"ftth_finland",[107,500,49,236,108,42],32,[188,502,40,375,189,42],33,[504,502,40,372,88,42],"40830554",[506,507,49,461,508,42],"40807677",34,"preciousoo",[510,511,34,333,512,42],"40819345",35,"Borborygymus",[],[515,520,525,530,534,539],{"id":516,"name":517,"points":518,"category":519},"40740021","rahulilango.com",1054,"learning_science",{"id":521,"name":522,"points":523,"category":524},"40789672","flipdisc",455,"creative",{"id":526,"name":527,"points":528,"category":529},"40757951","Web Check",411,"dev_tools",{"id":531,"name":532,"points":533,"category":529},"40789353","From dotenv to dotenvx",354,{"id":535,"name":536,"points":537,"category":538},"40802557","alpha.reconquer.online",340,"games",{"id":540,"name":541,"points":542,"category":543},"40788648","Triplit",279,"data",false,"v2",{"id":547,"name":5,"line":6,"ai":544,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":30,"top":18,"kw":548,"maker":11,"domain":10,"tier":19,"attempts":42,"strip":549,"url":9,"mix":550,"praise":100,"objection":128,"sub":21},40805010,[24,25,26,27,28],"5000005531514315500535503514343335533553",[32,33,34,35],1794036656857]