Umuvuduko wihuse hamwe na TTS idahwitse TTS
FastSpeech itanga imvugo yose yerekanwe muburyo bubangikanye aho kuba icyarimwe icyarimwe, bigatuma synthesis yihuta cyane kandi ihamye.
Incamake
It solved the slow, error-prone generation that plagued earlier autoregressive models like Tacotron.
Kwibira cyane
Moderi yambere ya TTS nka Tacotron 2 irigenga: barahanura buri cyiciro cyamajwi cyateganijwe kuribanjirije, kikaba gitinda kandi gikunda gusimbuka cyangwa gusubiramo amagambo mugihe ibitekerezo bidahwitse. FastSpeech, yatangijwe na Microsoft na kaminuza ya Zhejiang muri 2019, ihindura ibi mu guhanura icyarimwe icyarimwe. Umuyoboro-shimikiro-ugaburira umuyoboro ufata fonema, uhanura neza igihe buri fonema igomba kumara hamwe nuburebure burebure, kandi ikagura urutonde rwumubare ukwiye wamakadiri mbere yo kubyara spekrogramu mumurongo umwe. FastSpeech 2 yateye imbere kuri ibi mu guhanura ikibanza n'imbaraga, ndetse no guhugura intego zigihe cyo kuva ku gahato aho kubitandukanya nicyitegererezo cyabarimu gahoro, bitanga imvugo karemano kandi ishobora kugenzurwa.
Ubushishozi
Amayeri y'ingenzi ni uburebure bugenzura. Kuberako inyandiko n'amajwi bifite uburebure butandukanye, FastSpeech ihanura igihe kuri buri fonema kandi igasubiramo gusa iyo fonema yihishe inshuro nyinshi kugirango ihuze uburebure bwa spekrogramu. Uku guhuza kugaragara gusimbuza ibitekerezo byoroshye. Gukora buri kintu muburyo bubangikanye bisobanura igihe cyo kugereranya biterwa nuburebure bwinteruro, kandi gukuraho autoregressive loop bikuraho amakosa ya caskadi yo gusimbuka no gusubiramo ijambo.
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.
Ejo hazaza hihuta kandi byihuse TTS
Sintezike itari autoregressive ubu niyo isanzwe itanga umusaruro TTS kuko irihuta, ikomeye, kandi irashobora kugenzurwa. Sisitemu yigihe kizaza igana inzira nziza yo kugenzura, kugabanuka-gutinda gutambuka kubisabwa bizima, hamwe nimpinduka zanyuma-zanyuma zisimbuka hagati ya spekrogramu hagati. Diffusion- kandi itemba ishingiye kubintu bitari autoregressive moderi nayo irazamuka, ikomatanya ibisa na FastSpeech hamwe nubwiza bukomeye bwo kubyara, mugihe ikibanza cyihariye hamwe nigihe cyo kugenzura bikomeza guhabwa agaciro kubicuruzwa byijwi byahinduwe.
Gushyira mu bikorwa Isi
Porogaramu nyayo yo kugendana igihe-itanga amajwi-ihinduranya amajwi ako kanya ukoresheje parallel yihuta-yuburyo bwa synthesis.
Sisitemu y'abakiriya-serivisi IVR ihindura inyandiko yingirakamaro kumvugo ku munzani nta makosa yo gusimbuka.
Isomero rya ecran yabasomyi batanga imvugo yihuse, yizewe kubwinyandiko ndende kubikoresho byoroheje.
Ibikoresho bikubiyemo amajwi reka abarema bahindure ikibanza nigipimo cyo kuvuga mu buryo butaziguye, tubikesha ikibuga cyihuta cya FastSpeech 2 hamwe nabahanuzi b'ingufu.
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 FastSpeech and Non-Autoregressive TTS 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
Tortoise TTS Autoregressive Synthesis
Ibibazo bikunze kubazwa
What is FastSpeech and Non-Autoregressive TTS?
FastSpeech itanga imvugo yose yerekanwe muburyo bubangikanye aho kuba icyarimwe icyarimwe, bigatuma synthesis yihuta cyane kandi ihamye. Byakemuye ibisekuru bitinda, bikunda kwibasirwa nicyitegererezo cyambere autoregressive nka Tacotron.
Bisobanura iki 'non-autoregressive' bisobanura murwego rwa FastSpeech?
Non-autoregressive bisobanura amakadiri yakozwe muburyo bubangikanye mumurongo umwe, ntabwo ateganijwe umwe-umwe kumurongo wabanjirije.
Ni ikihe kibazo cya moderi ya autoregressive nka Tacotron 2 yihuta cyane?
Autoregressive attention irashobora kudahuza, igatera amagambo yasimbuwe cyangwa yasubiwemo, kandi decoding ikurikirana iratinda; Umuvuduko wihuse ukosora byombi.
Ni uruhe ruhare rw 'uburebure bugenzura' muri FastSpeech?
Uburebure bugenzura kwagura ibiranga foneme mugihe cyateganijwe, guhuza inyandiko ngufi ikurikiranye na ecran ndende.
Niki FastSpeech 2 yongeyeho ugereranije na FastSpeech yumwimerere?
FastSpeech 2 ihanura ikibanza n'imbaraga kandi ikoresha igihe-cyo guhuza igihe aho kuba umwarimu utinda, kunoza kamere no kugenzura.
Ni ukubera iki igihe cyo kwihuta cya FastSpeech gikura gusa nuburebure bwinteruro?
Kuberako ibisekuruza bisa, kongeramo amakadiri menshi ntabwo agwiza intambwe zikurikiranye nkuko decode ya autoregressive ikora.