UnivNet Multi-Resolution Vocoder
UnivNet shine vocoder na GAN wanda ke yin hukunci don ƙirƙirar sauti ta amfani da ƙididdiga masu yawa da aka lissafta a ƙudurin STFT daban-daban, yana ƙara dalla-dalla dalla-dalla.
Dubawa
It aims to be a universal vocoder that generalizes well to unseen speakers and recording conditions.
Zurfafa nutsewa
UnivNet, wanda Jang et al. a cikin 2021, yana magance rauni na gama-gari ga masu amfani da murya na GAN: manyan mitoci masu ɗorewa ko kayan tarihi. Yanayin janareta a kan cikakken band mel-spectrograms kuma yana amfani da jujjuyawar yanayi (LVC), inda ake hasashen kernels na jujjuyawar akan tashi daga fasalulluka na shigarwa don haka tace ta dace da abun cikin gida. Ra'ayin kanun labarai shine mai nuna wariya mai ƙuduri mai yawa (MRSD): maimakon yin hukunci da ɗanyen waveform kawai, UnivNet tana ƙididdige STFT da yawa tare da girman taga daban-daban da girman hop kuma yana gudanar da masu nuna wariya akan waɗannan girman. Wannan yana tura janareta don samun cikakkun bayanai masu kyau da faffadan tsarin ɗan lokaci daidai. An horar da masu magana da yawa, UnivNet yana samar da maganganun yanayi don muryoyin da ba ta taɓa gani ba yayin horo, yana samun lakabin duniya.
Fahimtar Fasaha
Juyin yanayi na UnivNet yana haifar da ma'aunin kernel ɗin sa a hankali daga yanayin yanayin sanyi ta hanyar ƙaramin hanyar sadarwa ta kernel-predictor, don haka kowane lokaci mataki yana amfani da ingantaccen tace abun ciki maimakon tsayayyen kwaya. Haɗe tare da wariya mai ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙayyadaddun ƙa'idodi, wanda ke ɗaukar yawancin cinikin-lokaci-lokaci lokaci guda, wannan kai tsaye yana kai hari ga babban rukunin mitar inda mafi sauƙi GAN vocoders ke yin blur ko hum.
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 UnivNet Multi-Resolution Vocoder
UnivNet's Multi-resolution spectrogram nuna bambanci ya zama daidaitaccen sinadari a cikin tarin TTS na zamani da tsarin tasiri kamar BigVGAN da codecs audio na jijiya. Yi tsammanin ƙirar duniya, mai magana-agnostic don ci gaba da faɗaɗa zuwa ga rera murya, haɗin harsuna da yawa, da cikakken sauti na 48 kHz, yayin da ra'ayin daidaitacce-kernel yana sanar da ingantattun samfuran na'urori waɗanda dole ne su sarrafa muryoyi daban-daban ba tare da daidaitawar kowane mai magana ba.
Aiwatar da Gaskiyar Duniya
Sabis na TTS masu magana da yawa waɗanda dole ne su yi sauti na halitta akan muryoyin da ba su cikin bayanan horo
Bututun muryar murya inda guda ɗaya vocoder na duniya ke hidima ga masu magana da yawa
Littafin jiwuwa mai inganci da ba da labari na kwasfan fayiloli yana buƙatar tsantsan sibilance da manyan mitoci
Vocoder na baya don tsarin TTS na ƙarshe-zuwa-ƙarshe waɗanda ke haɗa na'urar hangen nesa tare da janareta mai ƙarfi mai ƙarfi.
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 UnivNet Multi-Resolution Vocoder 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
Neural Vocoders
Tambayoyin da ake yawan yi
What is UnivNet Multi-Resolution Vocoder?
UnivNet shine vocoder na GAN wanda ke yin hukunci don ƙirƙirar sauti ta amfani da ƙididdiga masu yawa da aka lissafta a ƙudurin STFT daban-daban, yana ƙara dalla-dalla dalla-dalla. Yana da nufin zama vocoder na duniya wanda ke keɓanta da kyau ga masu magana da ba a gani da yanayin rikodi.
Menene fasalin sa hannun mai nuna wariya na UnivNet?
UnivNet's Multi-Resolution spectrogram discriminator yana nazarin STFTs da yawa tare da taga daban-daban da girman hop don kama nau'ikan cinikin lokaci-lokaci daban-daban.
Wace matsala ce a cikin GAN vocoders na farko UnivNet musamman ke niyya?
Ta hanyar nuna wariya a cikin ƙudirin bakan gizo-gizo da yawa, UnivNet yana haɓaka dalla-dalla dalla-dalla wanda mafi sauƙi na GAN vocoders sau da yawa blush.
Menene rikice-rikice masu canzawa-wuri (LVC) a cikin UnivNet?
LVC tana amfani da hanyar sadarwa ta kernel-predictor don haka kowane wuri yana amfani da matatun abubuwan da suka dace da abun ciki waɗanda aka samo daga yanayin mel-spectrogram.
Me yasa aka bayyana UnivNet a matsayin mai vocoder na duniya?
An horar da masu magana da yawa, UnivNet yana haɗa magana ta dabi'a har ma da muryoyin da ba ta taɓa haɗuwa da su ba a cikin horo, don haka duniya baki ɗaya.
Ta yaya bambance-bambancen spectrogram masu ƙuduri da yawa ke taimakawa janareta?
Kudirin STFT daban-daban suna jaddada ɓangarorin ciniki na lokaci-lokaci daban-daban, don haka daidaita dukkan su yana tura janareta zuwa cikakken daki-daki da tsari.