[{"data":1,"prerenderedAt":523},["ShallowReactive",2],{"project-39176570":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":50,"post_text":52,"what":53,"said":103,"maker_in_thread":216,"shape":218,"grid":223,"attempts":484,"same_week":485,"maker_shows":46,"maker_hidden":515,"reader_version":516,"card":517},"39176570","WhisperFusion","Low-latency conversations with an AI chatbot","title","Show HN: WhisperFusion – Low-latency conversations with an AI chatbot","https:\u002F\u002Fgithub.com\u002Fcollabora\u002FWhisperFusion","github.com","mfilion",1706538190,2024,272,102,252,3515,8,"legend","ai","voice assistant","other",[24,25,26,27,28],"speech recognition","low latency","voice chatbot","whisper","text to speech","curious",0.124,{"positive":32,"neutral":33,"mixed":34,"critical":35},13,46,4,6,69,{"design":38,"idea":41,"performance":43,"execution":45,"pricing":47,"docs":48,"privacy":49},{"positive":39,"critical":40},1,0,{"positive":42,"critical":40},7,{"positive":44,"critical":34},3,{"positive":39,"critical":46},2,{"positive":40,"critical":40},{"positive":40,"critical":40},{"positive":39,"critical":40},[44,44,51,44,44,44,44,40,51,34,44,44,44,44,39,40,34,44,51,40,40,40,44,34,39,44,39,44,44,34,44,51,40,44,51,40,39,51,51,44],5,"WhisperFusion builds upon the capabilities of open source tools WhisperLive and WhisperSpeech to provide a seamless conversations with an AI chatbot.",{"problem":54,"audience":66,"pricing":66,"motivation":66,"time_to_build":66,"stack":67,"openness":101,"stage":102,"team_size":102},{"value":55,"source":56,"evidence":57},"Slow, unnatural turn-taking in voice conversations with AI chatbots","inferred",[58,62],{"id":59,"by":60,"text":61},"39179316","localhost","There are two things that I think are needed and that I'm not sure if anyone provides yet to make this scenario work well: 1. Interruption - I need to be able to say \"hang on\"…",{"id":63,"by":64,"text":65},"39178457","pizzathyme","Whenever I walk my dog I find myself wanting a conversationalist LLM layer to exist in the best form. LLM's now are great at conversation, but the connective tissue between the…",null,[68,78,82,94],{"value":69,"source":70,"evidence":71},"Whisper (via WhisperLive)","maker",[72,74],{"id":73},"post",{"id":75,"by":76,"text":77},"39178257","renus","WhisperLive builds upon the Whisper model; for the demo, we used small.en, but you can also use large without introducing a bigger latency for the overall pipeline since the…",{"value":79,"source":70,"evidence":80},"WhisperSpeech",[81],{"id":73},{"value":83,"source":84,"evidence":85},"TensorRT","commenter",[86,90],{"id":87,"by":88,"text":89},"39180974","stiffler01","WhisperLive supports both TensorRT and faster-whisper. We didn’t reduce the chunk size rather use padding based on the chunk size received from the client. Reducing the segment…",{"id":91,"by":92,"text":93},"39181930","codethief","Seeing that this uses TensorRT (i.e. seems well optimized), what GPUs are supported? Could I run this on a Jetson?",{"value":95,"source":84,"evidence":96},"Mistral LLM",[97],{"id":98,"by":99,"text":100},"39177827","pilotneko","This project is using Mistral, not Phi-2. However, it is clear from reading the README.MD that this runs locally, so your point still stands. That being said, it looks like all…","open_source","unknown",{"praise":104,"objections":123,"requests":152,"questions":175,"alternatives":194},[105,116],{"value":106,"source":84,"evidence":107},"Impressive low-latency voice interaction",[108,112],{"id":109,"by":110,"text":111},"39179105","ramon156","Great to hear its seamless real-time ultra low-latency. Hopefully the next iteration is blazingly fast too!",{"id":113,"by":114,"text":115},"39177587","asynchronous","Very neat capability. Need to see more of hyper-optimizing models to one specific use case, this is a great example of doing so.",{"value":117,"source":84,"evidence":118},"Excellent packaging and project",[119],{"id":120,"by":121,"text":122},"39179638","doctorpangloss","This is an excellent project with excellent packaging. It is primarily a packaging problem. Why does every Python application on GitHub have its own ad-hoc, informally specified,…",[124,138,145],{"value":125,"source":84,"evidence":126},"Whisper is not designed for streaming; latency and WER unclear",[127,131,135],{"id":128,"by":129,"text":130},"39177708","regularfry","This is using phi-2, so the first assumption would be that it's local. It's a tiny little model in the grand scheme of things. I've been toying around with something similar…",{"id":132,"by":133,"text":134},"39178068","albertzeyer","See also the blog post: https:\u002F\u002Fwww.collabora.com\u002Fnews-and-blog\u002Fnews-and-events\u002Fwhisperfusion-ultra-low-latency-conversations-with-an-ai-chatbot.html WhisperFusion, WhisperLive,…",{"id":136,"by":133,"text":137},"39182931","Ah, but that sounds like a very inefficient approach, which probably still has quite high latency, and probably also performs bad in terms of word-error-rate (WER). But I'm happy…",{"value":139,"source":84,"evidence":140},"Lacks interruption support",[141,142],{"id":59,"by":60,"text":61},{"id":143,"by":88,"text":144},"39179629","Indeed a great point. Waiting for a specific cue, before responding, is an interesting idea. It would make the interaction more natural, especially in situations where the user…",{"value":146,"source":84,"evidence":147},"Hard to install for ordinary users",[148],{"id":149,"by":150,"text":151},"39179782","abdullahkhalids","We see these tools such as this posted several times a week. Is there any expectation they will be installable by the common person? Where is the setup.exe, .deb, .rpm, .dmg?",[153,160,167],{"value":154,"source":84,"evidence":155},"Interruption support",[156,157],{"id":59,"by":60,"text":61},{"id":158,"by":60,"text":159},"39179865","Thanks! Really looking forward to interruptions. I think about the cue as kind of being like \"Hey Siri\u002FAlexa\u002FCortana\" but in reverse.",{"value":161,"source":84,"evidence":162},"Latency and WER benchmark numbers",[163,164],{"id":132,"by":133,"text":134},{"id":165,"by":133,"text":166},"39178438","Yes, but when you change Whisper to make it live, to get WhisperLive, surely this has an effect on the WER, it will get worse. The question is, how much worse? And what is the…",{"value":168,"source":84,"evidence":169},"Cue-based turn taking",[170,171],{"id":59,"by":60,"text":61},{"id":172,"by":173,"text":174},"39179652","Valgrim","In order to feel like a human, cues should not be a pre-programmed phrase, the system should continuously listen to the conversation, and evaluate constantly if speaking is…",[176,183,187],{"value":177,"source":84,"evidence":178},"Does it run locally or call remote APIs?",[179],{"id":180,"by":181,"text":182},"39182370","WhackyIdeas","I had a quick read through, maybe I missed something but is this all local run or does it need api access to OpenAI’s remote system? The reason I ask is that I’m building…",{"value":184,"source":84,"evidence":185},"Which GPUs are supported, can it run on a Jetson?",[186],{"id":91,"by":92,"text":93},{"value":188,"source":84,"evidence":189},"Can TensorRT be replaced with CoreML?",[190],{"id":191,"by":192,"text":193},"39185841","jdkee","Does anyone know if you can replace TensorRT with a similar call to Apple's CoreML for the same functionality?",[195,202,209],{"value":196,"source":84,"evidence":197},"Vocode",[198],{"id":199,"by":200,"text":201},"39179078","monkeydust","This post reminded me of Vocode: https:\u002F\u002Fgithub.com\u002Fvocodedev\u002Fvocode-python Discussion on them here from 10 months ago: https:\u002F\u002Fnews.ycombinator.com\u002Fitem?id=35358873 I tried the…",{"value":203,"source":84,"evidence":204},"Kaldi streaming ASR models",[205],{"id":206,"by":207,"text":208},"39186157","huac","whisper is simply not designed for this, in many ways, and it's impressive engineering to try and overcome its limitations, but I can't help but feel that it is easier to just…",{"value":210,"source":84,"evidence":211},"Persona \u002F SmarterChild voice demos",[212],{"id":213,"by":214,"text":215},"39180237","lambdaba","> Interruption Well, today is your lucky day!: https:\u002F\u002Fpersona-webapp-beta.vercel.app\u002F and the demo https:\u002F\u002Fsmarterchild.chat\u002F",{"comments":40,"answered":40,"key":217},[],{"commenters":219,"max_depth":220,"maker_replies":40,"first_comment_mins":221,"lifespan_hours":222},60,9,83,68,[224,226,230,234,237,240,242,244,245,247,250,253,256,260,263,267,268,271,275,278,281,283,287,290,294,298,301,305,308,311,312,314,315,318,320,321,323,324,327,329,332,335,339,340,343,344,347,349,352,355,357,360,362,365,368,370,372,374,377,380,383,386,387,390,393,394,396,398,400,403,404,406,409,411,414,416,418,421,424,427,430,433,435,437,439,441,444,447,449,452,455,458,460,462,464,467,469,471,474,476,479,481],[59,40,44,225,60,40],"project\u002Fneutral\u002Fdesign",[227,40,44,228,229,40],"39179563","project\u002Fneutral\u002Fidea","bilsbie",[231,40,40,232,233,40],"39179606","other\u002Fneutral\u002Fnone","andai",[235,40,40,232,236,40],"39179587","plufz",[238,40,44,239,76,40],"39180497","project\u002Fneutral\u002Fperformance",[143,40,44,241,88,40],"maker\u002Fneutral\u002Fidea",[158,40,51,243,60,40],"maker\u002Fpositive\u002Fidea",[172,40,44,228,173,40],[246,40,44,228,60,40],"39179841",[248,40,40,232,249,40],"39179730","philsnow",[251,40,44,225,252,40],"39180347","dr_kiszonka",[254,40,40,232,255,40],"39184781","m463",[257,40,44,258,259,40],"39179812","project\u002Fneutral\u002Fexecution","pksebben",[261,40,44,262,76,40],"39180367","maker\u002Fneutral\u002Fexecution",[264,40,44,265,266,40],"39179846","project\u002Fmixed\u002Fexecution","zan2434",[213,40,44,228,214,40],[269,40,44,239,270,40],"39181736","dmw_ng",[272,40,40,273,274,40],"39184270","commenter\u002Fneutral\u002Fnone","selcuka",[276,40,40,273,277,40],"39184486","meandmycode",[279,40,40,273,280,40],"39185158","lolinder",[282,40,44,258,274,40],"39187042",[284,40,51,285,286,40],"39182172","project\u002Fpositive\u002Fperformance","napier",[288,40,44,289,214,40],"39182850","project\u002Fneutral\u002Fpricing",[291,40,44,292,293,40],"39182449","project\u002Fmixed\u002Fperformance","spywaregorilla",[295,40,51,296,297,40],"39187177","project\u002Fpositive\u002Fdesign","derwiki",[299,40,51,285,300,40],"39186092","erhaetherth",[302,40,39,303,304,40],"39189051","project\u002Fcritical\u002Fexecution","jbkkd",[306,40,44,258,307,40],"39214344","3371",[309,40,44,228,310,40],"39181927","irthomasthomas",[132,39,44,239,133,40],[75,39,44,313,76,40],"maker\u002Fneutral\u002Fperformance",[165,39,44,239,133,40],[316,39,44,317,76,40],"39178612","project\u002Fneutral\u002Fdocs",[319,39,44,317,133,40],"39178682",[87,39,44,313,88,40],[136,39,39,322,133,40],"project\u002Fcritical\u002Fperformance",[206,39,39,322,207,40],[325,39,40,232,326,40],"39188482","Oranguru",[328,39,40,232,133,40],"39188781",[330,39,40,232,331,40],"39178695","082236036778",[333,39,44,334,76,40],"39178809","maker\u002Fneutral\u002Fdocs",[336,39,51,337,338,40],"39178472","project\u002Fpositive\u002Fidea","pyryt",[120,46,44,265,121,40],[341,46,40,232,342,40],"39179713","traverseda",[199,44,44,228,200,40],[345,44,40,232,346,40],"39179777","domrdy",[348,44,40,232,200,40],"39180646",[350,34,51,337,351,40],"39177737","cristyansv",[353,34,40,232,354,40],"39177974","column",[356,34,40,232,351,40],"39178803",[358,34,40,232,359,40],"39178944","summarity",[361,34,40,232,351,40],"39179140",[363,34,40,232,364,40],"39179525","fragmede",[366,51,40,232,367,40],"39178998","lxe",[369,51,40,232,270,40],"39179290",[371,51,40,232,364,40],"39179460",[373,51,40,232,270,40],"39179632",[375,35,44,239,376,40],"39179021","wruza",[378,35,44,239,379,40],"39179495","freeqaz",[381,35,44,382,76,40],"39179693","maker\u002Fneutral\u002Fnone",[384,42,51,337,385,40],"39177615","quonn",[128,42,39,322,129,40],[388,42,44,239,389,40],"39177765","gpderetta",[391,42,40,392,207,40],"39179709","commenter\u002Fneutral\u002Fperformance",[98,42,44,239,99,40],[395,42,40,273,129,40],"39178414",[397,42,44,313,76,40],"39178352",[399,42,39,322,129,40],"39178475",[401,42,44,239,402,40],"39178562","intalentive",[91,18,44,258,92,40],[405,18,44,262,88,40],"39187017",[407,220,44,228,408,40],"39177669","yieldcrv",[410,220,44,228,338,40],"39178433",[412,220,44,228,413,40],"39179034","zachthewf",[415,220,44,241,88,40],"39179259",[417,220,51,243,88,40],"39179240",[419,420,44,228,338,40],"39178491",10,[180,422,44,423,181,40],11,"project\u002Fneutral\u002Fprivacy",[425,422,44,426,76,40],"39182448","maker\u002Fneutral\u002Fprivacy",[428,422,51,429,181,40],"39182571","maker\u002Fpositive\u002Fprivacy",[431,432,51,337,229,40],"39179588",12,[434,432,40,273,233,40],"39179620",[436,432,40,273,229,40],"39179664",[438,432,40,232,233,40],"39192572",[440,432,44,423,259,40],"39179736",[442,432,40,443,229,40],"39180101","other\u002Fneutral\u002Fprivacy",[445,432,40,443,446,40],"39182533","blooalien",[448,432,51,337,214,40],"39180281",[450,32,44,228,451,40],"39178683","tasty_freeze",[453,32,40,232,454,40],"39178737","jsheard",[456,32,40,273,457,40],"39179192","realo",[459,32,40,273,454,40],"39179454",[149,461,39,303,150,40],14,[463,461,44,334,76,40],"39180623",[113,465,51,466,114,40],15,"project\u002Fpositive\u002Fexecution",[191,468,44,258,192,40],16,[109,470,51,285,110,40],17,[63,472,40,473,64,40],18,"other\u002Fneutral\u002Fidea",[475,472,44,241,76,40],"39178583",[477,472,44,265,478,40],"39178680","hombre_fatal",[480,472,44,241,76,40],"39178936",[482,472,44,228,483,40],"39178836","visarga",[],[486,491,496,501,506,510],{"id":487,"name":488,"points":489,"category":490},"39180953","Boardzilla",574,"languages",{"id":492,"name":493,"points":494,"category":495},"39159418","Pong Wars",490,"games",{"id":497,"name":498,"points":499,"category":500},"39191870","Ubicloud",475,"infrastructure",{"id":502,"name":503,"points":504,"category":505},"39148336","PlatoEdu",289,"learning_science",{"id":507,"name":508,"points":509,"category":20},"39201182","gpt",280,{"id":511,"name":512,"points":513,"category":514},"39207570","Stanchion",273,"data",false,"v2",{"id":518,"name":5,"line":6,"ai":515,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":519,"top":18,"kw":520,"maker":11,"domain":10,"tier":19,"attempts":40,"strip":521,"url":9,"mix":522,"praise":106,"objection":125,"sub":21},39176570,0.12,[24,25,26,27,28],"3353333054333310435000341313343503501553",[32,33,34,35],1793991083662]