I-VITS Ukuhlanganiswa Kwenkulumo Yokuphela-kuya-Ekupheleni
I-VITS imodeli yokuguqula umbhalo ube inkulumo eshintsha umbhalo ngokuqondile uwenze amaza omsindo angahluziwe kusistimu eyodwa eqeqeshiwe, yeqa ipayipi elivamile lezigaba ezimbili.
Uhlolojikelele
By combining variational inference with adversarial training, it produces remarkably natural, expressive speech.
I-Deep Dive
I-VITS (I-Variational Inference ehambisana nokufunda kokuphikisana kombhalo-to-ekupheleni wombhalo-kuya-enkulumweni), owethulwe nguKim, Kong, kanye neNdodana ngo-2021, ihlanganisa imibono emithathu amasistimu amadala ayigcina ehlukene. I-autoencoder enemibandela enemibandela (VAE) ifunda ukumelwa okucashile kwenkulumo, ukugeleza okuvamile kwenza lokho kusatshalaliswa okucashile kuvumelana nezimo ngokwanele ukuze kuthwebule imininingwane emihle ye-acoustic, futhi umbandlululi wesitayela se-GAN uphusha i-waveform ekhiqiziwe iye kumaqiniso. Okubaluleke kakhulu, i-VITS iqeqesha imodeli ye-acoustic kanye nevokhoda ndawonye kunezigaba ezimbili, isusa ukungafani okwehlisa izinga lapho amamojula eqeqeshwa ngokuhlukana. Futhi yethula isibikezelo sobude besikhathi se-stochastic, ukuze umusho ofanayo ukhulunywe ngezigqi ezihlukene, ezizwakalayo zemvelo isikhathi ngasinye.
I-Technical Insight
I-VITS ixazulula inkinga yokuqondanisa ngokusesha kwe-Monotonic Alignment (MAS), ethola imephu ehamba phambili phakathi kwamathokheni ombhalo namafreyimu omsindo phakathi nokuqeqeshwa ngaphandle kokuqondanisa kwangaphandle. Ingemuva le-VAE libalwa ngekhompyutha lisuka kumsindo wangempela, kuyilapho okubekwe ngaphambilini embhalweni kuphinde kumiswe ngokugeleza okujwayelekile ukuze kufane nawo. Uma kucatshangelwa, wenza isampula embhalweni ngaphambi kwalokho bese unquma ngokuqondile ku-waveform, ngakho-ke ayikho i-mel-spectrogram ehlukile futhi ayikho i-vocoder ehlukile edingekayo.
I-Strategic Impact
Finyelela futhi ufinyelele
Ithuthukisa ukufinyeleleka ngokuloba, ukulandisa, nezixhumi ezibonakalayo zezwi.
Izindleko kanye nesabelomali
Amaqembu emidiya angathumela umsindo opholishiwe ngokushesha ngamabhajethi amancane.
Isivinini nesikali
Amasistimu abhekene nekhasimende angacubungula ukusebenzelana okukhulunyiwe ngesilinganiso esikhulu.
Ikusasa le-VITS End-to-End Speech Synthesis
I-VITS idale umndeni wabalandelayo obusa umthombo ovulekile we-TTS. I-VITS2 yenza ukwakhiwa kwaba lula kanye nemvelo ethuthukisiwe, kuyilapho i-YourTTS kanye ne-Coqui XTTS esetshenziswa kabanzi yanweba indlela yokuhlanganisa izwi nezilimi eziningi. Lindela ukuqhubeka komsebenzi ezintweni ezilula, zesikhathi sangempela ezikudivayisi, ukumbozwa okungcono kwezilimi eziningi ngezilimi ezisetshenziswa kancane, nokulawula okuqinile emizweni nesitayela sokukhuluma, njengoba idizayini esuka ekupheleni iyisisekelo esikhangayo, esiqondwa kahle ongakhela kuso.
Ukuqaliswa Komhlaba Wangempela
I-Coqui TTS ithumela amamodeli asuselwa ku-VITS abathuthukisi abawashuna kahle ukuze bahlanganise izwi lomlandi othile kuma-audiobook.
Abasizi bezwi bomthombo ovulekile kuzingxenyekazi zekhompuyutha ze-Raspberry Pi-class basebenzisa amamodeli e-VITS ahlangene ukuze bakhiphe inkulumo engaxhunyiwe ku-inthanethi ngokugcwele.
Izinhlelo zokusebenza zokufunda ulimi zenza izibonelo zokuphimisa zemvelo zisebenzisa okuhlukile kwe-VITS yezilimi eziningi njenge-YourTTS.
Izitudiyo zegeyimu ye-Indie zihlanganisa imigqa yengxoxo ye-NPC eyahlukahlukene, incike ekubikezeleni ubude besikhathi se-stochastic ngesigqi esingelona irobhothi.
Izingozi & Guardrails
Ukusetshenziswa kabi kwezwi kanye nezingozi zokuzenza ongeyena ziyanda uma imvume ingekho.
Ukunemba kungase kwehle kuzo zonke izinhlobo zokuphimisela, izilimi zesigodi, noma izindawo ezinomsindo.
Umsindo wokwenziwa ungenziwa iphutha njengenkulumo eyiqiniso ngaphandle kokulebula okucacile.
Ukuqalisa Umhlahlandlela
Thola imvume esobala yokuthwebula izwi, ukuhlanganisa, nokusebenzisa kabusha.
Ikhwalithi yokuhlola kuzo zonke izipikha nezimo zangemuva.
Chaza ukuthi kunini lapho umuntu kufanele abuyekeze noma agunyaze okuphumayo.
Lebula umsindo wokwenziwa futhi ugcine amarekhodi atholakalayo ukuze aziphendulele.
Qhubeka Uhlole
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 VITS End-to-End Speech Synthesis 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
Umhlahlandlela olandelayo
I-Emotional Speech Synthesis
Imibuzo evame ukubuzwa
What is VITS End-to-End Speech Synthesis?
I-VITS imodeli yokuguqula umbhalo ube inkulumo eshintsha umbhalo ngokuqondile uwenze amaza omsindo angahluziwe kusistimu eyodwa eqeqeshiwe, yeqa ipayipi elivamile lezigaba ezimbili. Ngokuhlanganisa ukucatshangelwa okuguquguqukayo nokuqeqeshwa kokuphikisana, kukhiqiza inkulumo engokwemvelo ngokuphawulekayo, evezayo.
Imelelani i-acronym VITS?
I-VITS imele i-Variational Inference ehambisana nokufunda okuphikisayo kombhalo-ube-ukuphela kwenkulumo, okubonisa izithako zayo ezintathu eziyinhloko.
Uyini umehluko omkhulu wezakhiwo phakathi kwe-VITS namapayipi endabuko e-TTS?
I-VITS isukela ekupheleni: ifunda ifomethi yombhalo uye-wave kumodeli eyodwa, igwema ukungafani kwemodeli ehlukile ye-acoustic nevokhoda.
I-VITS ikufunda kanjani ukuqondanisa phakathi kombhalo nozimele bomsindo?
I-VITS isebenzisa i-Monotonic Alignment Search ukuze ithole ukuqondanisa okungcono kakhulu kombhalo nohlaka ngaphakathi, ngaphandle kokuqondanisa kwangaphandle okudingekayo.
Kungani i-VITS ifaka isibikezelo sobude besikhathi se-stochastic?
Isibikezelo sobude besikhathi se-stochastic sivumela umusho ofanayo ukhulunywe ngezigqi zemvelo ezihlukene, kugwenywe ukugeleza okuyisicaba, okurobhothi.
Iyiphi ingxenye ephusha okukhiphayo kwe-VITS kumsindo ozwakalayo wangempela?
I-VITS isebenzisa ukuqeqeshwa kwe-adversarial (GAN), lapho umbandlululi ehlulela ukuthi amagagasi azwakala angokoqobo, athuthukisa ikhwalithi yokucabanga.