Daidaici WaveGAN Vocoder
Parallel WaveGAN shine muryar murya mai sauri wanda ke juya mel-spectrogram zuwa sigar sautin sauti mai ɗanɗano ta amfani da ƙaramin GAN, yana samar da duk samfuran lokaci ɗaya.
Dubawa
It matters because it gives near-real-time, high-quality speech with a compact model.
Zurfafa nutsewa
Vocoder shine mataki na ƙarshe na bututun TTS: yana canza taswirar fasalin sauti (yawanci mel-spectrogram) zuwa ainihin motsin sautin da kuke ji. Parallel WaveGAN, wanda Yamamoto, Song, da Kim suka gabatar a cikin 2019, yana yin wannan tare da janareta salon WaveNet mara-autoregressive wanda aka horar dashi azaman hanyar sadarwa ta gaba. Maimakon tsinkayar samfurin sauti guda ɗaya a lokaci guda kamar WaveNet na asali, yana samar da dukkan nau'ikan igiyoyin ruwa a layi daya, yana sa shi sauri da sauri. Maɓallin girke-girkensa ya haɗu da asarar gaba tare da asarar gajeriyar juzu'i mai yawa Fourier transform (STFT), don haka ƙirar ta dace da ainihin sigina a cikin lokuta da yawa da ma'auni. Sakamakon shine ƙaramin janareta (kusan sigogi miliyan 1.4) wanda ke tafiyar da sauri sau da yawa fiye da ainihin lokacin akan GPU.
Fahimtar Fasaha
Janareta babbar hanyar sadarwa ce mai faɗaɗa-convolution da aka sharadi akan mel-spectrogram da shigar da amo, ƙarar taswira da fasali kai tsaye zuwa samfurori. Horon da aka haɗa tare yana rage girman asarar STFT mai ƙididdigewa, ƙididdigewa ta hanyar kwatanta girman sikirin mai girma da tsayin FFT da yawa, da asarar gaba daga mai nuna wariya yana yanke hukunci na gaskiya. Kalmar STFT tana daidaitawa da haɓaka horo na gaba, yana ɗaukar cikakkun bayanai dalla-dalla da faffadan siffa ba tare da distillation ba.
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 Parallel WaveGAN Vocoder
Parallel WaveGAN ya taimaka wajen kafa GAN vocoders a matsayin tsoho mai amfani, kuma asarar STFT mai ƙuduri mai yawa a yanzu yana bayyana a cikin magada kamar HiFi-GAN da tsarin yawo da yawa. Halin yanayin yana nuni zuwa ga ƙarami, ƙananan latency vocoders don mataimakan na'ura, na'urorin ji, da sauya murya mai rai, tare da vocoders na duniya waɗanda ke gabaɗaya ga masu magana da ba a gani. Yi tsammanin haɗin kai tare da ƙarshen-zuwa-ƙarshen TTS da ingantaccen turawa akan wayar hannu da kwakwalwan kwamfuta da aka saka.
Aiwatar da Gaskiyar Duniya
Fitowar magana ta ainihi a cikin mataimakan muryar wayar hannu inda latency da girman ƙirar ke da mahimmanci
Yin hidima azaman janareta na waveform haɗe tare da samfuran sauti kamar Tacotron 2 ko FastSpeech
Rubutu-zuwa-magana akan na'urar don kayan aikin samun dama waɗanda ba za su iya dogara ga gajimare ba
Tsarin jujjuya murya wanda ke sake haɗa jujjuyawar spectrograms zuwa sauti mai sauti na halitta
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 Parallel WaveGAN 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
MelGAN Generative Vocoder
Tambayoyin da ake yawan yi
What is Parallel WaveGAN Vocoder?
Parallel WaveGAN shine muryar murya mai sauri wanda ke juya mel-spectrogram zuwa sigar sautin sauti mai ɗanɗano ta amfani da ƙaramin GAN, yana samar da duk samfuran lokaci ɗaya. Yana da mahimmanci saboda yana ba da kusan-ainihin lokaci, magana mai inganci tare da ƙaramin tsari.
Menene aikin mai vocoder kamar Parallel WaveGAN?
Vocoder shine matakin TTS na ƙarshe wanda ke haɗa ainihin raƙuman sauti daga fasalulluka na sauti kamar mel-spectrogram.
Ta yaya Parallel WaveGAN ke samar da samfuran sauti idan aka kwatanta da WaveNet na asali?
Parallel WaveGAN ba autoregressive ba ne, yana samar da dukkan nau'in igiyar ruwa a lokaci ɗaya maimakon samfurin ta samfurin, wanda ya sa ya fi sauri.
Wace asara ce ke tsakiyar Parallel WaveGAN banda asarar gaba?
Ya haɗu da asarar gaba tare da asarar STFT mai ƙuduri mai yawa wanda ke kwatanta girman sikeli a ma'auni da yawa.
Wane gine-ginen Parallel WaveGAN janareta ke amfani da shi?
Janareta cibiyar sadarwa ce mai kama da WaveNet da aka gina daga faɗuwar juzu'i, mai sharadi akan mel-spectrogram da amo.
Me yasa Parallel WaveGAN ke da kyau don turawa?
Tare da kusan sigogi miliyan 1.4 yana ƙarami kuma yana tafiyar da sauri sau da yawa fiye da ainihin lokacin, manufa don amfani akan na'urar.