Iringaniza WaveGAN Vocoder
Parallel WaveGAN nijwi ryihuta ryijwi rihindura mel-spectrogramu muburyo bwamajwi mbisi ikoresheje GAN ntoya, ikabyara icyarimwe icyarimwe.
Incamake
It matters because it gives near-real-time, high-quality speech with a compact model.
Kwibira cyane
Ijwi ni icyiciro cya nyuma cyumuyoboro wa TTS: ihindura ikarita yerekana imiterere ya acoustic (mubisanzwe mel-spectrogramu) mumajwi nyirizina wumva. Parallel WaveGAN, yatanzwe na Yamamoto, Indirimbo, na Kim muri 2019, ibikora hamwe na generator yo mu bwoko bwa WaveNet idahwitse yatojwe nkumuyoboro uhanganye. Aho guhanura icyitegererezo kimwe cyamajwi icyarimwe nka WaveNet yumwimerere, itanga imiterere yose yumurongo ugereranije, bigatuma yihuta cyane. Ibyingenzi byingenzi bihuza igihombo cyo guhangana nigisubizo cyigihe gito cyo guhinduka kwa Fourier (STFT), bityo moderi ihuye nikimenyetso nyacyo mugihe kinini nubunzani. Igisubizo ni generator ntoya (ibipimo bya miriyoni 1,4) ikora inshuro nyinshi kurenza igihe nyacyo kuri GPU.
Ubushishozi
Imashini itanga umuyoboro mugari wagutse kuri mel-spectrogramu no kwinjiza urusaku, gushushanya urusaku hiyongereyeho ibimenyetso byerekana ingero. Amahugurwa afatanya kugabanya igihombo kinini cya STFT igihombo, ubaze ugereranije ubunini bwa spekrogramu yubunini bwa FFT nuburebure bwa hop, hamwe nigihombo cyo guhangana nivangura rishingiye ku kuri. Ijambo STFT rihamye kandi ryihutisha imyitozo yo guhangana, ifata ibisobanuro byiza byose hamwe nuburyo bugari butagaragara.
Ingaruka z'Ingamba
Kugera no kugera
Itezimbere kugerwaho binyuze mu kwandukura, kuvuga, no guhuza amajwi.
Igiciro na bije
Amatsinda yibitangazamakuru arashobora kohereza amajwi yihuse hamwe na bije nto.
Umuvuduko n'igipimo
Sisitemu ireba abakiriya irashobora gutunganya imikoranire ivugwa murwego runini.
Kazoza Kuringaniza WaveGAN Vocoder
Parallel WaveGAN yafashije gushiraho amajwi ya GAN nkibisanzwe bifatika, kandi igihombo cyayo kinini-STFT igihombo kigaragara mubasimbuye nka HiFi-GAN hamwe na sisitemu nyinshi zitemba. Inzira nyabagendwa yerekeza kuri buri gihe ntoya, amajwi-yihuta yo gutabaza kubafasha ku bikoresho, ibyuma bifata amajwi, hamwe no guhindura amajwi bizima, hiyongereyeho amajwi rusange asanzwe avuga abavuga batagaragara. Witegereze guhuza hamwe na TTS iherezo-iherezo no kohereza neza kuri mobile igendanwa.
Gushyira mu bikorwa Isi
Ijambo-nyaryo risohoka mumajwi igendanwa abafasha aho ubukererwe nubunini bwikitegererezo bifite akamaro
Gukora nka generator ya generator ihujwe na moderi ya acoustic nka Tacotron 2 cyangwa FastSpeech
Ku gikoresho-cy-inyandiko-ku-bikoresho kubikoresho byoroshye bidashobora kwishingikiriza ku gicu
Sisitemu yo guhindura amajwi yongeye guhinduranya spekrogramu mu majwi asanzwe yumvikana
Ingaruka & Kurinda
Gukoresha nabi amajwi no kwigira ibyago byiyongera mugihe uruhushya rubuze.
Ukuri kurashobora kugabanuka hejuru yimvugo, imvugo, cyangwa urusaku rwibidukikije.
Amajwi yubukorikori arashobora kwibeshya kumvugo yukuri nta kirango gisobanutse.
Igishushanyo mbonera
Shaka uruhushya rusobanutse rwo gufata amajwi, gukoroniza, no gukoresha.
Ikizamini cyiza mubiganiro bitandukanye hamwe nuburyo bwimbere.
Sobanura igihe umuntu agomba gusuzuma cyangwa kwemeza ibisubizo.
Andika amajwi yubukorikori kandi ugumane inyandiko zerekana kubazwa.
Komeza Ubushakashatsi
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
Ubuyobozi bukurikira
MelGAN Vocoder
Ibibazo bikunze kubazwa
What is Parallel WaveGAN Vocoder?
Parallel WaveGAN nijwi ryihuta ryijwi rihindura mel-spectrogramu muburyo bwamajwi mbisi ikoresheje GAN ntoya, ikabyara icyarimwe icyarimwe. Nibyingenzi kuko itanga hafi-nyayo-nyayo, ireme-mvugo nziza hamwe na moderi yoroheje.
Nuwuhe murimo wijwi nka Parallel WaveGAN?
Ijwi ni icyiciro cya nyuma cya TTS gihuza amajwi nyirizina avuye muri acoustic nka mel-spectrogram.
Nigute Parallel WaveGAN itanga urugero rwamajwi ugereranije na WaveNet yumwimerere?
Kuringaniza WaveGAN ntabwo ari autoregressive, itanga icyerekezo cyose icyarimwe aho kuba icyitegererezo nicyitegererezo, bigatuma cyihuta cyane.
Ni ikihe gihombo ari ingenzi kuri Parallel WaveGAN usibye igihombo cyo guhangana?
Ihuza igihombo cyo guhangana nigihombo kinini cya STFT igereranya ubunini bwa spekrogramu kumunzani.
Ni ubuhe bwubatsi generator ya Parallel WaveGAN ikoresha?
Imashini itanga umuyoboro umeze nka WaveNet wubatswe kuva kwaguka, kugengwa na mel-spectrogramu n urusaku.
Ni ukubera iki Parallel WaveGAN ishimishije kubohereza?
Hafi ya miriyoni 1,4 ni ntoya kandi ikora inshuro nyinshi byihuse kuruta igihe nyacyo, cyiza cyo gukoresha ibikoresho.