Maganar Halitta da Latent Yawa TTS
NaturalSpeech layin bincike ne na Microsoft TTS da ke nufin ingancin magana ta matakin ɗan adam, tare da sigar baya ta amfani da ɓoyayyen ɓoye don samar da wadatattun muryoyin halitta.
Dubawa
It shows how diffusion models, famous for images, can produce expressive, controllable audio.
Zurfafa nutsewa
Asalin Magana na NaturalSpeech (2022) shine tsarin farko da aka bayar da rahoton isa ga ingancin matakin ɗan adam akan maƙasudin LJSpeech, waɗanda masu sauraro suka yanke hukunci waɗanda ba za su iya dogaro da su ba daga rikodin rikodi na gaske. Ya yi amfani da bambance-bambancen autoencoder tare da madaidaitan abubuwan da suka gabata don rufe rata tsakanin horo da ƙima. NaturalSpeech 2 sannan ya ɗauki hanyar watsawa ta latent: magana tana ɓoye ta hanyar codec mai jiwuwa na jijiyoyi zuwa cikin ci gaba da ɓoyayyiyar ɓoyayyiyar ɓarna, kuma ƙirar watsawa tana koyo don samar da waɗancan latents daga rubutu, yana ba da damar ƙarar sautin murya mai ƙarfi daga ɗan gajeren lokaci. NaturalSpeech 3 ya gabatar da yaduwa mai ma'ana, raba magana zuwa halaye marasa daidaituwa kamar abun ciki, prosody, timbre, da dalla-dalla, don haka kowane za'a iya ƙirƙira da sarrafa kansa don aminci da sassauci.
Fahimtar Fasaha
Rushewar ɓoye yana aiki ta ƙara amo zuwa ƙarami na latent wakilci na magana da horar da hanyar sadarwa don juyar da hayaniyar mataki-mataki. Maimakon yin watsi da raƙuman raƙuman ruwa ko cikakkun bayanai, NaturalSpeech 2 yana musanta latent codec, waɗanda ƙananan-girma ne kuma masu sauƙin ƙira. Ƙaddamar da rubutu da muryar magana tana sadar da jujjuyawar, don haka samfurin latent na ƙarshe da aka ƙirƙira zuwa magana wanda ya dace da abun ciki da ake buƙata da ainihin mai magana.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar Maganar Halitta da Latent Diffusion TTS
Tushen watsawa da madaidaitan TTS suna nuni zuwa muryoyin da ba na halitta bane kawai amma masu kyawu, barin masu amfani su daidaita timbre, motsin rai, da haɓaka azaman bugun kira masu zaman kansu. Yi tsammanin samfura cikin sauri ta hanyar distillation da ƴan-mataki watsawa, mafi ƙarfi cloning sifili daga dakika na audio, da kuma matsananci hadewa tare da manyan harshe model domin sanin mahallin. Waɗannan ci gaban kuma suna ƙara buƙatar alamar ruwa da kuma kariyar yarda, tun da babban amintaccen cloning yana haifar da haɗarin rashin amfani.
Aiwatar da Gaskiyar Duniya
Dubbing Studios suna haɗa muryar ɗan wasan kwaikwayo daga ɗan gajeren samfurin don gano fina-finai, ta amfani da nau'in nau'in nau'in nau'in nau'in nau'in nau'in nau'i na NaturalSpeech 2.
Kafofin watsa labarai na audio suna haifar da labari-matakin mutum wanda masu sauraro ke fafutukar bambancewa da basirar murya ta gaske.
Kayan aikin samun dama suna sake ƙirƙirar muryar mutum daga tsoffin rikodi don waɗanda suka rasa magana.
Rukunin ƙirƙirar abun ciki suna barin masu gyara su daidaita timbre da ƙwaƙƙwaran ƙima, suna ba da damar sifofin ƙirƙira na NaturalSpeech 3.
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the NaturalSpeech and Latent Diffusion TTS quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Jagora na gaba
Stable Audio Latent Diffusion
Tambayoyin da ake yawan yi
What is NaturalSpeech and Latent Diffusion TTS?
NaturalSpeech layin bincike ne na Microsoft TTS da ke nufin ingancin magana ta matakin ɗan adam, tare da sigar baya ta amfani da ɓoyayyen ɓoye don samar da wadatattun muryoyin halitta. Yana nuna yadda samfuran watsawa, shahararrun hotuna, ke iya samar da sauti mai bayyanawa, mai iya sarrafawa.
Wane mataki ne aka bayar da rahoton cim ma ainihin Maganar NaturalSpeech (2022)?
An bayar da rahoton NaturalSpeech a matsayin tsarin farko don isa ga ingancin matakin ɗan adam akan LJSpeech, tare da masu sauraro ba su iya dogara da su bambanta da ainihin magana.
Menene samfurin yaɗuwar ɓoye na NaturalSpeech 2 a zahiri ke haifarwa?
NaturalSpeech 2 yana bazuwa akan latent codec, waɗanda ke ƙanƙanta da sauƙi don ƙira fiye da ɗan raƙuman raƙuman ruwa, sannan yana yanke su zuwa sauti.
Ta yaya samfurin watsawa ke samar da bayanai a babban matakin?
Yadawa yana ƙara hayaniya yayin horo kuma ya koyi juyar da shi, yana samar da samfurori ta hanyar ci gaba da ƙi daga hayaniyar bazuwar.
Wanne maɓalli na iyawar ɓoyewar ɓoye ke ba da NaturalSpeech 2?
Ta hanyar daidaitawa a kan gajeriyar muryar faɗakarwa, NaturalSpeech 2 na iya haɗa muryar mai magana da ba a taɓa horar da ita a kai ba.
Menene babban ra'ayin NaturalSpeech 3's 'factorized diffusion'?
Factorized diffusion disentangles, prosody, timbre, and acoustic daki-daki don haka kowane sifa za a iya ƙirƙira da sarrafa shi daban.