[{"data":1,"prerenderedAt":684},["ShallowReactive",2],{"project-46832074":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":46,"post_text":50,"what":51,"said":79,"maker_in_thread":200,"shape":212,"grid":217,"attempts":645,"same_week":646,"maker_shows":49,"maker_hidden":676,"reader_version":677,"card":678},"46832074","ear","I trained a 9M speech model to fix my Mandarin tones","title","Show HN: I trained a 9M speech model to fix my Mandarin tones","https:\u002F\u002Fsimedw.com\u002F2026\u002F01\u002F31\u002Fear-pronunication-via-ctc\u002F","simedw.com","simedw",1769820687,2026,469,153,60,4366,2,"legend","languages","pronunciation feedback","web_app",[24,25,26,27],"mandarin tones","pronunciation","speech recognition","language learning","mixed",0.128,{"positive":31,"neutral":32,"mixed":33,"critical":34},18,21,7,13,59,{"execution":37,"performance":39,"idea":41,"design":44,"docs":45},{"positive":38,"critical":18},0,{"positive":18,"critical":40},10,{"positive":42,"critical":43},12,1,{"positive":43,"critical":38},{"positive":38,"critical":38},[47,38,47,43,18,47,38,43,38,38,38,38,38,38,38,38,43,47,38,47,48,43,43,43,48,47,38,43,48,48,18,49,43,48,43,47,48,48,48,47],3,5,4,"Built this because tones are killing my spoken Mandarin and I can't reliably hear my own mistakes.\n\nIt's a 9M Conformer-CTC model trained on ~300h (AISHELL + Primewords), quantized to INT8 (11 MB), runs 100% in-browser via ONNX Runtime Web.\n\nGrades per-syllable pronunciation + tones with Viterbi forced alignment.\n\nTry it here: https:\u002F\u002Fsimedw.com\u002Fprojects\u002Fear\u002F",{"problem":52,"audience":58,"pricing":63,"motivation":64,"time_to_build":63,"stack":68,"openness":77,"stage":78,"team_size":77},{"value":53,"source":54,"evidence":55},"Hard to hear one's own Mandarin tone mistakes","maker",[56],{"id":57},"post",{"value":59,"source":60,"evidence":61},"Mandarin learners","inferred",[62],{"id":57},null,{"value":65,"source":54,"evidence":66},"Tones are killing their spoken Mandarin and they cannot reliably hear their own mistakes",[67],{"id":57},[69,73],{"value":70,"source":54,"evidence":71},"Conformer-CTC",[72],{"id":57},{"value":74,"source":54,"evidence":75},"ONNX Runtime Web",[76],{"id":57},"unknown","beta",{"praise":80,"objections":104,"requests":142,"questions":173,"alternatives":192},[81,93],{"value":82,"source":83,"evidence":84},"Intuitive UI and quick demo","commenter",[85,89],{"id":86,"by":87,"text":88},"46832799","nirvanatikku","talk about 30 seconds to wow. great app, UX and demo. would love to use this. kudos.",{"id":90,"by":91,"text":92},"46832424","rahimnathwani","This is incredible. When I was first learning Chinese (casually, ~20 years ago), my teacher used some Windows software that drew a diagram of the shape of my pronunciation, so…",{"value":94,"source":83,"evidence":95},"Useful for learners without native speakers nearby",[96,100],{"id":97,"by":98,"text":99},"46832506","drekipus","instantly awesome. I suck at chinese but I want to get better and I'm too embarassed to try and talk with real people and practise. This is a great compromise. even just…",{"id":101,"by":102,"text":103},"46838659","ctkhn","This is fantastic. Been looking for a way to get feedback on my pronunciation since I came back from Shanghai and haven't been seeing native speakers every day. Is there any plan…",[105,116,131],{"value":106,"source":83,"evidence":107},"Misrecognizes tones in fast or conversational speech",[108,112],{"id":109,"by":110,"text":111},"46832680","dapangzi","Longtime lurker, made an account specifically to give feedback here as an intermediate speaker. :) This is a great initiative and I hope to see more come out of this; I am not…",{"id":113,"by":114,"text":115},"46833316","tifan","Well, it would work only when I speak word by word, not as a sentence or in a normal speed for daily conversations. The model thinks I was making mistakes when I speak casually…",{"value":117,"source":83,"evidence":118},"Sensitive to background noise and mic quality",[119,123,127],{"id":120,"by":121,"text":122},"46836264","alixwang","As a native speaker of Mandarin the demo it's not work for me. It can't check the pronounce of my voice. I don't know what's wrong of it, may be it's too sensitive(my daughter…",{"id":124,"by":125,"text":126},"46833923","olalonde","It might be a mic issue but my wife, who is a native speaker, seems to get most characters wrong. I will try again later in a quieter place to see if that helps.",{"id":128,"by":129,"text":130},"46834199","wenjian","Chinese here, some of the tune is wrong, maybe the env here has some noise, good luck on learning mandarin ;)",{"value":132,"source":83,"evidence":133},"Flags correct native pronunciation as wrong",[134,138],{"id":135,"by":136,"text":137},"46832590","affogarty","This is extremely cool, although I asked my wife (who is Chinese) to try it out and it said she made some mistakes.",{"id":139,"by":140,"text":141},"46834840","redleader55","This is a very cool to have! Thanks for putting the time to build it. For me it doesn't work very well. Even easy phrases like 他很忙 get transcribed completely random \"ma he yu\".…",[143,158,169],{"value":144,"source":83,"evidence":145},"Support for other languages",[146,150,154],{"id":147,"by":148,"text":149},"46837622","alexandermorgan","I wish this were available for more languages! It would also be neat to estimate the native language of the speaker, given their pronunciation of the target language, and propose…",{"id":151,"by":152,"text":153},"46834034","martianlantern","Nice! I need something similar for english now",{"id":155,"by":156,"text":157},"46832342","jellojello","This is amazing, if you feel like opening an entire language to being learned more easily.. Farsi is a VERY overlooked language, my wife\u002Fher family speak it but it's so difficult…",{"value":159,"source":83,"evidence":160},"Release inference and training code",[161,165],{"id":162,"by":163,"text":164},"46839912","jainaayush05","Any plans on releasing the inference\u002Ftraining code?",{"id":166,"by":167,"text":168},"46834421","mentalgear","Very cool ! Will you make the source available as well?",{"value":170,"source":83,"evidence":171},"Desktop or mobile version",[172],{"id":101,"by":102,"text":103},[174,185],{"value":175,"source":83,"evidence":176},"How the model was trained and on what data",[177,181],{"id":178,"by":179,"text":180},"46835650","kris_builds","Super interesting project. Curious about the data collection - did you record yourself, use existing datasets, or both? I've been thinking about building something similar for…",{"id":182,"by":183,"text":184},"46835619","victorbjorklund","Cool. Would love a write up about how you did it if you have time",{"value":186,"source":83,"evidence":187},"Whether Cantonese is supported",[188],{"id":189,"by":190,"text":191},"46833074","stuxnet79","How difficult would it be to adapt this to Cantonese? It is a surprisingly difficult language to learn. It has more tones than Mandarin plus comparatively less access to learning…",[193],{"value":194,"source":83,"evidence":195},"Azure Speech Studio",[196],{"id":197,"by":198,"text":199},"46833615","cocoa19","Have you tried the Azure Speech Studio? I wonder how your custom model compares to this solution. I played around with python scripts for the same purpose. The AI gives feedback…",{"comments":201,"answered":43,"key":202},6,[203,206,209],{"id":204,"text":205},"46841928","Thank for the great feedback! I have just added sandhi support, please let me know if it's working better.",{"id":207,"text":208},"46841636","Hi, thanks for the feedback. The 了 issue was a bug on the JavaScript side; that should be fixed (training did thankfully handle it correctly). The other two are probably things…",{"id":210,"text":211},"46832486","For accents, I’ve mostly tested with a few friends so far. I’m wondering whether region should be a parameter, because training on all dialects might make the system too lax.",{"commenters":213,"max_depth":214,"maker_replies":201,"first_comment_mins":215,"lifespan_hours":216},101,8,34,116,[218,220,223,227,230,232,234,236,239,241,244,247,250,253,257,259,262,264,267,270,272,275,278,281,284,287,290,292,295,298,301,304,306,307,310,313,316,319,321,324,327,330,332,334,337,340,343,346,348,350,352,355,358,360,362,364,366,368,370,372,375,377,380,382,383,386,388,390,393,395,398,401,404,406,409,412,415,417,419,421,425,428,431,434,437,439,442,444,447,450,453,455,457,459,463,466,469,472,475,477,479,481,483,486,488,490,495,499,501,503,505,509,511,512,514,518,522,524,528,530,532,536,540,544,548,553,555,559,563,566,570,574,578,582,584,587,589,592,594,598,600,603,606,608,612,616,619,623,627,631,633,637,641],[109,38,47,219,110,38],"project\u002Fmixed\u002Fexecution",[221,38,38,222,114,38],"46833331","commenter\u002Fneutral\u002Fnone",[224,38,38,225,226,38],"46833680","other\u002Fneutral\u002Fnone","et-al",[228,38,38,225,229,38],"46833769","dirteater_",[231,38,38,225,110,38],"46837092",[233,38,38,225,91,38],"46837938",[235,38,38,225,110,38],"46839803",[237,38,38,225,238,38],"46834154","allan_s",[240,38,38,225,110,38],"46840743",[242,38,47,219,243,38],"46833345","sqs",[245,38,38,222,246,38],"46833831","jhanschoo",[248,38,38,222,249,38],"46834454","jimz",[251,38,38,222,252,38],"46834414","carlmr",[254,38,43,255,256,38],"46833498","project\u002Fcritical\u002Fexecution","mercanlIl",[204,38,38,258,11,43],"maker\u002Fneutral\u002Fexecution",[260,38,47,261,110,38],"46843701","project\u002Fneutral\u002Fexecution",[263,38,47,219,110,38],"46850107",[265,43,43,255,266,38],"46834396","yunusabd",[268,43,47,219,269,38],"46834831","vjerancrnjak",[271,43,38,222,266,38],"46834913",[273,43,38,222,274,38],"46843143","qingcharles",[276,18,47,219,277,38],"46834826","namelosw",[279,18,38,225,280,38],"46834893","samiv",[282,18,38,225,283,38],"46835024","zelphirkalt",[285,18,38,225,286,38],"46834937","mijoharas",[288,18,38,225,289,38],"46834974","tianqi",[291,18,38,222,277,38],"46840667",[293,18,38,225,294,38],"46841258","calf",[296,18,38,222,297,38],"46841749","thaumasiotes",[299,18,43,300,283,38],"46835000","project\u002Fcritical\u002Fperformance",[302,18,38,222,303,38],"46836638","samus",[305,18,38,225,277,38],"46864874",[207,18,38,258,11,43],[308,47,38,225,309,38],"46832695","ecshafer",[311,47,38,225,312,38],"46832780","cyberax",[314,47,38,225,315,38],"46833899","barrell",[317,47,38,225,318,38],"46834686","thenthenthen",[320,47,38,225,312,38],"46850027",[322,47,38,225,323,38],"46832948","dionian",[325,47,38,225,326,38],"46832964","laurieg",[328,47,38,225,329,38],"46833921","ronyeh",[331,47,38,225,297,38],"46849853",[333,47,38,225,329,38],"46891377",[335,47,38,225,336,38],"46833210","danparsonson",[338,47,38,225,339,38],"46833622","bunderbunder",[341,47,38,225,342,38],"46833302","vjvjvjvjghv",[344,47,38,225,345,38],"46834581","DiogenesKynikos",[347,47,38,222,303,38],"46836717",[349,47,38,222,345,38],"46837210",[351,47,38,225,345,38],"46834508",[353,47,38,225,354,38],"46835105","snicky",[356,47,38,222,357,38],"46840392","Sxubas",[359,47,38,222,345,38],"46844201",[361,47,38,222,294,38],"46841405",[363,47,38,222,345,38],"46844216",[365,47,38,222,297,38],"46849889",[367,47,38,222,345,38],"46850316",[369,47,38,222,297,38],"46851128",[371,47,38,222,345,38],"46871171",[373,47,38,222,374,38],"46857763","dtm987654123",[376,47,38,222,345,38],"46865456",[378,47,38,222,379,38],"46840075","cvhc",[381,47,38,222,357,38],"46840314",[113,49,43,300,114,38],[384,48,38,225,385,38],"46832387","vunderba",[210,48,38,387,11,43],"maker\u002Fneutral\u002Fnone",[389,48,47,387,385,38],"46834255",[391,48,38,225,392,38],"46832675","zdragnar",[394,48,38,222,309,38],"46832685",[396,48,38,225,397,38],"46834171","mleonhard",[399,48,38,225,400,38],"46832955","luckydata",[402,48,38,225,403,38],"46832736","devin",[405,48,38,225,312,38],"46832787",[407,48,38,225,408,38],"46833552","sowbug",[410,201,47,411,339,38],"46833551","project\u002Fmixed\u002Fidea",[413,201,38,222,414,38],"46833715","zdc1",[416,201,38,225,315,38],"46833992",[90,33,48,418,91,38],"project\u002Fpositive\u002Fperformance",[420,33,38,225,266,38],"46834426",[422,214,47,423,424,38],"46833386","project\u002Fneutral\u002Fdesign","memalign",[426,214,43,300,427,38],"46833393","data_ders",[429,214,47,423,430,38],"46834775","siwatanejo",[432,214,38,222,433,38],"46834778","knocte",[435,214,38,436,11,43],"46841940","maker\u002Fneutral\u002Fdesign",[120,438,43,300,121,38],9,[440,438,38,441,11,43],"46836823","maker\u002Fneutral\u002Fperformance",[135,40,47,443,136,38],"project\u002Fmixed\u002Fperformance",[445,40,43,300,446,38],"46833688","hawflakes",[448,449,38,225,283,38],"46835094",11,[451,449,38,225,452,38],"46835166","peterburkimsher",[454,449,38,225,283,38],"46836016",[456,449,47,387,452,38],"46840412",[458,449,38,225,283,38],"46836107",[460,42,48,461,462,38],"46833648","project\u002Fpositive\u002Fidea","frozennothing",[464,34,38,225,465,38],"46833006","ChadNauseam",[189,467,47,468,190,38],14,"project\u002Fneutral\u002Fidea",[470,467,38,225,471,38],"46834062","inkyoto",[473,467,38,225,474,38],"46834423","hnfong",[476,467,38,222,294,38],"46841568",[478,467,38,222,474,38],"46847842",[480,467,38,222,471,38],"46843649",[482,467,38,222,474,38],"46847735",[484,467,38,222,485,38],"46853289","nathanhammond",[487,467,38,225,474,38],"46853600",[489,467,38,225,485,38],"46855790",[491,492,43,493,494,38],"46833327",15,"project\u002Fcritical\u002Fidea","rablackburn",[496,497,47,387,498,38],"46837063",16,"sgt",[197,500,47,387,198,38],17,[502,500,38,225,229,38],"46833811",[97,31,48,504,98,38],"project\u002Fpositive\u002Fnone",[506,507,48,461,508,38],"46837394",19,"erdemo",[101,510,48,504,102,38],20,[178,32,47,387,179,38],[147,513,47,468,148,38],22,[515,516,38,225,517,38],"46832976",23,"SequoiaHope",[519,520,47,423,521,38],"46834092",24,"arjie",[139,523,43,300,140,38],25,[525,526,47,468,527,38],"46835505",26,"tomaytotomato",[124,529,43,300,125,38],27,[162,531,47,387,163,38],28,[533,534,48,461,535,38],"46832972",29,"byb",[537,538,48,461,539,38],"46832925",30,"jrockway",[541,542,38,225,543,38],"46834328",31,"sim04ful",[545,546,38,225,547,38],"46833155",32,"baby",[549,550,47,551,552,38],"46834467",33,"project\u002Fneutral\u002Fperformance","JCharante",[554,215,43,300,323,38],"46832934",[556,557,48,418,558,38],"46855110",35,"holg",[560,561,48,504,562,38],"46832724",36,"bytesandbits",[564,561,38,225,565,38],"46832768","anonzzzies",[567,568,48,461,569,38],"46837582",37,"namr2000",[571,572,47,423,573,38],"46837834",38,"while_true_",[575,576,43,300,577,38],"46838326",39,"eudamoniac",[579,580,38,225,581,38],"46832576",40,"btrlsnqtn",[583,580,38,225,238,38],"46834064",[182,585,47,586,183,38],41,"maker\u002Fneutral\u002Fdocs",[588,585,47,586,179,38],"46835657",[166,590,47,591,167,38],42,"project\u002Fneutral\u002Fdocs",[151,593,48,461,152,38],43,[595,596,48,461,597,38],"46841017",44,"irl_zebra",[155,599,48,461,156,38],45,[601,599,38,602,11,43],"46832427","maker\u002Fneutral\u002Fidea",[604,599,38,225,605,38],"46832549","kranner",[607,599,38,225,452,38],"46835232",[609,610,48,461,611,38],"46836365",46,"maximedupre",[613,614,47,551,615,38],"46840880",47,"cheonn638",[86,617,48,618,87,38],48,"project\u002Fpositive\u002Fdesign",[620,621,48,461,622,38],"46834239",49,"felixbecker",[624,625,47,468,626,38],"46842759",50,"somesun",[628,629,48,461,630,38],"46832797",51,"cmuguythrow",[128,632,43,300,129,38],52,[634,635,47,468,636,38],"46845924",53,"kanwisher",[638,639,48,461,640,38],"46833261",54,"iamanllm",[642,643,38,225,644,38],"46835276",55,"contingencies",[],[647,652,657,662,667,672],{"id":648,"name":649,"points":650,"category":651},"46847690","Craftplan",577,"web_saas",{"id":653,"name":654,"points":655,"category":656},"46850205","NanoClaw",533,"ai",{"id":658,"name":659,"points":660,"category":661},"46850803","Xikipedia",441,"learning_science",{"id":663,"name":664,"points":665,"category":666},"46882389","Ghidra MCP Server",298,"dev_tools",{"id":668,"name":669,"points":670,"category":671},"46868479","Safe-now.live",195,"infrastructure",{"id":673,"name":674,"points":675,"category":671},"46824877","Amla Sandbox",146,false,"v2",{"id":679,"name":5,"line":6,"ai":676,"cat":20,"year":13,"points":14,"comments":15,"rec":28,"s":680,"top":18,"kw":681,"maker":11,"domain":10,"tier":19,"attempts":38,"strip":682,"url":9,"mix":683,"praise":82,"objection":106,"sub":21},46832074,0.13,[24,25,26,27],"3031230100000000130351115301552415135553",[31,32,33,34],1794036691857]