Audio AI JAGORA

WaveGlow-Tsarin Vocoder

WaveGlow shine vocoder na tushen jijiya daga NVIDIA wanda ke haɗa nau'ikan kalaman magana daga mel-spectrograms a cikin wucewa ɗaya ba tare da jujjuyawar kai ba.

2 min karatuAn sabunta ta ƙarshe

Dubawa

It matters because it delivers high-quality audio faster than real time using only a simple likelihood loss.

Zurfafa nutsewa

WaveGlow, wanda Prenger, Valle, da Catanzaro suka fitar a NVIDIA a cikin 2018, ya haɗu da ra'ayoyi daga Glow da WaveNet don gina vocoder wanda ke da sauri da sauƙin horarwa. Ba kamar GAN vocoders ba, yana gudana ne mai daidaitawa: yana koyon taswirar da ba za a iya jujjuyawa ba tsakanin sauƙin rarraba Gaussian da siginar sauti mai jiwuwa, mai sharadi akan mel-spectrogram. Horowa yana haɓaka ainihin yuwuwar log ɗin bayanai, don haka baya buƙatar wani keɓantaccen mai wariya, babu jujjuyawar kai-da-kai, kuma babu rarrabuwar kawuna na malami-dalibi na hanyar sadarwa guda biyu waɗanda a baya a layi daya hanyoyin WaveNet ke buƙata. Don samar da sauti kuna gwada hayaniyar Gaussian kuma ku gudanar da hanyar sadarwa mara jujjuyawa a baya. WaveGlow yana samar da magana mai inganci kwatankwacin WaveNet yayin da yake haɗawa da sauri fiye da ainihin lokaci akan GPU na zamani.

Fahimtar Fasaha

WaveGlow yana tara matakan kwarara marar jujjuyawa, kowanne yana haɗa layin haɗin gwiwa tare da jujjuyawar juzu'i 1x1 da aka aro daga Glow. Ana haɗa samfuran sauti zuwa vectors ta hanyar matsi don haka yadudduka na iya canza su da kyau. Domin kowane mataki baya juyewa, jagorar gaba tana ƙididdige yuwuwar horo da taswirorin juyar da amo zuwa sauti don tunani. Cibiyoyin sadarwa guda ɗaya da makasudin yiwuwar log ɗin mara kyau ɗaya sun sa horo ya zama mai karko da sauƙi.

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 WaveGlow Tushen Vocoder

WaveGlow ya nuna cewa masu amfani da sauti masu tsafta na iya yin hamayya da ingancin autoregressive, suna tasiri daga baya kwarara da samfuran sauti masu dacewa. Sauƙin sa na asarar guda ɗaya ya kasance mai ban sha'awa, kodayake GAN vocoders kamar HiFi-GAN yanzu galibi suna cin nasara akan girma da sauri. Neman gaba, tushen kwarara da ra'ayoyin da suka dace suna sake farfadowa a cikin watsawa na zamani-kusa da TTS, da kuma WaveGlow-style invertible designs ci gaba da sanar da bincike a kan ainihin-yiwuwar, sarrafawa, da ingantaccen tsarin igiyar ruwa.

Aiwatar da Gaskiyar Duniya

Haɗa tare da Tacotron 2 a cikin bututun TTS na NVIDIA don samar da ingantaccen magana mai inganci

Haɗawar magana mai sauri na GPU don ba da labari, ɗab'i, da ayyukan ƙirƙirar abun ciki

Samar da horo da sauti na demo a cikin bincike inda aka fi son kwanciyar hankali, horarwar asara guda ɗaya

Fitowar murya ta gaske-lokaci a cikin tsarin mu'amala da ke gudana akan kayan aikin NVIDIA

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

1

Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.

2

Gwajin ingantattun masu magana daban-daban da yanayin baya.

3

Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.

4

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 WaveGlow Flow-Based Vocoder quiz

Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.

Fara tambayoyi

Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation

Jagora na gaba

Haɓaka Ƙarfafa Maganar Akwatin Murya

Tambayoyin da ake yawan yi

What is WaveGlow Flow-Based Vocoder?

WaveGlow shine vocoder na tushen jijiya daga NVIDIA wanda ke haɗa nau'ikan kalaman magana daga mel-spectrograms a cikin wucewa ɗaya ba tare da jujjuyawar kai ba. Yana da mahimmanci saboda yana ba da sauti mai inganci cikin sauri fiye da ainihin lokacin ta amfani da hasara mai sauƙi kawai.

WaveGlow ya haɗu da ra'ayoyin gine-gine daga wane nau'i biyu?

WaveGlow yana haɗa tsarin gudana mara jujjuyawar Glow tare da ƙirar ƙirar sauti na WaveNet.

Wane irin samfuri ne WaveGlow?

WaveGlow gudana ne mai daidaitawa wanda ke koyon taswira mara jujjuyawa tsakanin amo da sauti na Gaussian.

Ta yaya WaveGlow ke haifar da sigar igiyar ruwa a lokacin ƙididdigewa?

Saboda hanyar sadarwa ba ta juyowa, kuna zana hayaniyar Gaussian kuma kuna tafiyar da gudu a baya, mai sharadi akan mel-spectrogram.

Wace manufa WaveGlow ke inganta yayin horo?

A matsayin gudana, WaveGlow yana haɓaka ainihin yuwuwar log, ba buƙatar nuna bambanci ko distillation.

Wadanne abubuwa biyu ne suka hada kowane mataki mara jujjuyawa a cikin WaveGlow?

Kowane mataki na kwarara yana haɗa nau'i-nau'i na affin haɗin gwiwa tare da juzu'in juzu'i 1x1 wanda ba zai iya jujjuyawa ba daga Glow.