[{"data":1,"prerenderedAt":291},["ShallowReactive",2],{"project-22107144":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":5,"maker":10,"posted_at":11,"year":12,"points":13,"comments":14,"rank_in_year":15,"shown_in_year":16,"top":17,"tier":18,"category":19,"subcategory":20,"format":21,"keywords":22,"reception":28,"sentiment":29,"stance_mix":30,"scored_comments":34,"aspects":35,"strip":40,"post_text":44,"what":45,"said":71,"maker_in_thread":159,"shape":166,"grid":169,"attempts":253,"same_week":254,"maker_shows":43,"maker_hidden":283,"reader_version":284,"card":285},"22107144","colab.research.google.com","Train a language model to talk like you","title","Show HN: Train a language model to talk like you","https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1iHcQ8_K0cfRE3v8QX6FMKAzdSSGtf5IX","MasterScrat",1579617022,2020,209,32,260,3028,9,"notable","ai","language model fine-tuning","other",[23,24,25,26,27],"gpt-2","chat logs","language model","fine-tuning","google colab","curious",-0.083,{"positive":31,"neutral":32,"mixed":33,"critical":31},4,10,0,18,{"execution":36,"idea":37,"privacy":39},{"positive":33,"critical":33},{"positive":38,"critical":31},2,{"positive":33,"critical":33},[33,41,42,41,42,42,43,33,41,33,43,41,33,42,33,41,41,33,43,41,41,33,41,33,33,33,33,33,33,33,43,41],3,1,5,null,{"problem":46,"audience":44,"pricing":44,"motivation":53,"time_to_build":44,"stack":57,"openness":70,"stage":70,"team_size":70},{"value":47,"source":48,"evidence":49},"Generating conversational text that imitates a specific person's chat style","maker",[50],{"id":51,"by":10,"text":52},"22107163","You may have seen my recent post about [Chatistics: a Python tool to parse your Messenger\u002FHangouts\u002FWhatsApp\u002FTelegram chat logs into…",{"value":54,"source":48,"evidence":55},"Notebook for an AMLD 2020 workshop, Meet your Artificial Self",[56],{"id":51,"by":10,"text":52},[58,62,66],{"value":59,"source":48,"evidence":60},"Python",[61],{"id":51,"by":10,"text":52},{"value":63,"source":48,"evidence":64},"GPT",[65],{"id":51,"by":10,"text":52},{"value":67,"source":48,"evidence":68},"Google Colab",[69],{"id":51,"by":10,"text":52},"unknown",{"praise":72,"objections":85,"requests":111,"questions":128,"alternatives":143},[73],{"value":74,"source":75,"evidence":76},"Fun demo and workshop material","commenter",[77,81],{"id":78,"by":79,"text":80},"22107921","heybrandons","Thanks for sharing MasterScrat! This looks fun!",{"id":82,"by":83,"text":84},"22112542","raidicy","That sounds great! This and the main workshop site have many subjects I'm really interested to check out. Thank you!",[86,101],{"value":87,"source":75,"evidence":88},"Title suggests speech, but it only generates text",[89,93,97],{"id":90,"by":91,"text":92},"22110555","capableweb","I got a bit tricked by the title here on HN. Maybe we can replace `talk` with `write`? Thought this was something that could learn how I speak and could generate sound from that,…",{"id":94,"by":95,"text":96},"22108643","arethuza","I'm disappointed that this is about typed text rather than actual talking - I had hoped that training something that talked like me might assist technology vendors in actually…",{"id":98,"by":99,"text":100},"22113427","moron4hire","Yeah, Microsoft has had NN-based speech generators that can mimick your own voice for about a year now. Thought this was going to be a competing service.",{"value":102,"source":75,"evidence":103},"Privacy concern about uploading chat logs to Google Colab",[104,108],{"id":105,"by":106,"text":107},"22107715","thisisastopsign","I’ve never used PyTorch before... is this running within my local machine, or is there some API in here that’s also sending data to Google to also train their models? Asking a…",{"id":109,"by":79,"text":110},"22107912","The python notebook is hosted on google colab which will execute on free (for you) google servers. If you’re concerned about privacy probably do not upload your personal chat…",[112,121],{"value":113,"source":75,"evidence":114},"Share implementation details and notebooks",[115,118],{"id":116,"by":10,"text":117},"22113277","I would be very curious to see your notebook - while this simple approach works well with GPT, we are not getting the results we'd want with a more complex question\u002Fanswer model…",{"id":119,"by":10,"text":120},"22112130","We don't have any plan to record it currently. But we will release the two other notebooks used during the workshop, and plan to write a blog post detailing the full content.",{"value":122,"source":75,"evidence":123},"Add voice generation",[124],{"id":125,"by":126,"text":127},"22110982","thedirt0115","How about this combined with recent advances in text-to-speech like Deep Voice that can sound like you? (Edit: punctuation)",[129,136],{"value":130,"source":75,"evidence":131},"Python 2 or 3 and whether it runs in local Jupyter",[132],{"id":133,"by":134,"text":135},"22113395","prophesi","I definitely need to give this a whirl. Does it use Python 2 or 3, and is it as simple as importing its ipynb to run it in a local Jupyter notebook?",{"value":137,"source":75,"evidence":138},"Whether it works with small Q&A\u002FFAQ corpora",[139],{"id":140,"by":141,"text":142},"22108410","fudged71","Could you train this on a Q&A\u002FFAQ corpus and get somewhat relevant results? (And is there any better tool for doing this?)",[144,152],{"value":145,"source":75,"evidence":146},"Lyrebird \u002F Microsoft voice cloning",[147,151],{"id":148,"by":149,"text":150},"22109114","the-dude","I totally missed that Lyrebird was acquired : https:\u002F\u002Fnews.ycombinator.com\u002Fitem?id=21006405",{"id":98,"by":99,"text":100},{"value":153,"source":75,"evidence":154},"PPLM plug-and-play language model",[155],{"id":156,"by":157,"text":158},"22109182","perturbation","This is cool - might be worth training a simple discriminator model to identify your utterances, and then you can use the plug-and-play language model (PPLM -…",{"comments":31,"answered":42,"key":160},[161,162,163],{"id":51,"text":52},{"id":116,"text":117},{"id":164,"text":165},"22126928","Awesome :D We've published the third notebook: https:\u002F\u002Fcolab.research.google.com\u002Fdrive\u002F1XYNef9zcHhTjt6kM6ydL9oXTshoRknIV We would gladly have your feedback on our approach",{"commenters":167,"max_depth":31,"maker_replies":31,"first_comment_mins":42,"lifespan_hours":168},26,46,[170,172,174,176,177,178,179,183,185,188,190,193,194,196,200,204,209,212,213,214,216,217,221,223,226,230,234,237,240,244,247,249],[51,33,33,171,10,42],"project\u002Fneutral\u002Fidea",[133,33,41,173,134,33],"project\u002Fneutral\u002Fexecution",[90,42,42,175,91,33],"project\u002Fcritical\u002Fidea",[125,42,41,171,126,33],[98,42,42,175,99,33],[94,38,42,175,95,33],[180,41,43,181,182,33],"22113174","project\u002Fpositive\u002Fidea","Tenoke",[116,41,33,184,10,42],"maker\u002Fneutral\u002Fexecution",[186,41,41,187,182,33],"22126150","maker\u002Fneutral\u002Fnone",[164,41,33,189,10,42],"maker\u002Fpositive\u002Fnone",[191,41,43,181,192,33],"22113449","hug",[156,31,41,171,157,33],[148,43,33,195,149,33],"other\u002Fneutral\u002Fnone",[197,198,42,175,199,33],"22107808",6,"data_ders",[201,198,33,202,203,33],"22110663","commenter\u002Fneutral\u002Fnone","ferCats99",[205,206,41,207,208,33],"22110387",7,"project\u002Fneutral\u002Fnone","bryanrasmussen",[210,211,41,207,83,33],"22111849",8,[119,211,33,187,10,42],[82,211,43,189,83,33],[105,17,41,215,106,33],"project\u002Fneutral\u002Fprivacy",[109,17,41,215,79,33],[218,32,33,219,220,33],"22117839","other\u002Fneutral\u002Fprivacy","woefulregret",[140,222,41,171,141,33],11,[224,222,33,195,225,33],"22110459","cyorir",[227,228,33,195,229,33],"22111818",12,"MadWombat",[231,232,33,195,233,33],"22107879",13,"alfonsodev",[235,232,33,195,236,33],"22110926","pjmorris",[238,232,33,195,239,33],"22111767","whatshisface",[241,242,33,195,243,33],"22110507",14,"nickster",[245,242,33,195,246,33],"22111731","neodymiumphish",[78,248,43,189,79,33],15,[250,251,41,171,252,33],"22108769",16,"brainzap",[],[255,260,265,269,274,278],{"id":256,"name":257,"points":258,"category":259},"22063744","No-Bullshit Games",520,"games",{"id":261,"name":262,"points":263,"category":264},"22073037","ESQLate",505,"dev_tools",{"id":266,"name":267,"points":268,"category":264},"22081419","OneDev",459,{"id":270,"name":271,"points":272,"category":273},"22091603","A set of standard document templates",433,"productivity",{"id":275,"name":276,"points":277,"category":273},"22065333","Screenshot Hero",429,{"id":279,"name":280,"points":281,"category":282},"22077052","sixty5o2",230,"hardware",false,"v2",{"id":286,"name":5,"line":6,"ai":283,"cat":19,"year":12,"points":13,"comments":14,"rec":28,"s":287,"top":17,"kw":288,"maker":10,"domain":5,"tier":18,"attempts":33,"strip":289,"url":9,"mix":290,"praise":74,"objection":87,"sub":20},22107144,-0.08,[23,24,25,26,27],"03131150305301033053303000000053",[31,32,33,31],1793991002408]