Inkomoko-Akayunguruzo Vocoding n'isi
Indangururamajwi nigikoresho gitandukanya imvugo mubice byubaka kandi ikongera ikubaka.
Incamake
The source-filter model and the WORLD vocoder are classic methods that power text-to-speech and voice conversion by separating what your vocal cords do from what your mouth shapes.
Kwibira cyane
Inkomoko-muyunguruzi yerekana imvugo nkibice bibiri bikorera hamwe: isoko (urusaku ruvuye mu majwi yawe yinyeganyeza yijwi ryijwi ryumvikana, cyangwa umwuka wuzuye urusaku rwongorerana) Ijwi risesengura amajwi yafashwe kugirango agereranye ibyo bice, hanyuma ashushanya amajwi mashya muri yo. ISI, yasohowe na Masanori Morise ahagana mu mwaka wa 2016, nijwi ryiza cyane rikuramo amajwi atatu: F0 (ikibuga cyerekana isoko), ibahasha yerekana (ibiyungurura, ikoresheje algorithm ya CheapTrick), hamwe na aperiodicity (uko urusaku rungana nijwi, binyuze kuri PLATINUM / D4C). Iyi migezi itatu irashobora guhindurwa yigenga hanyuma igasubirwamo, bigatuma ISI ikorera kumurimo wa TTS hamwe no kuririmba sisitemu yijwi.
Ubushishozi
Imbaraga z'isi ziva mu gutandukana neza. CheapTrick igereranya ibahasha yoroheje igaragara neza ku makosa mato ya F0, mugihe DIO / Harvest track ikibuga na D4C bipima aperiodicity. Kuberako ikibanza, timbre, hamwe n urusaku bibaho muburyo butandukanye, urashobora guhindura F0 hejuru ya octave udahinduye uwo ijwi risa, cyangwa kurambura igihe utabanje guhindura ikibanza. Indangururamajwi nka WaveNet yaje kwerekana imiterere yumurongo utaziguye, ariko ISI ikomeza kwihuta, gusobanurwa, kandi nta ruhushya.
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 h'isoko-Muyunguruzi Vocoding n'isi
Amajwi meza atunganya amajwi yarenganijwe cyane cyane nijwi ryamajwi (HiFi-GAN, WaveRNN) kubwimiterere-yohejuru, ariko ISI ntabwo yazimye. Irokoka nkumuvuduko wihuse, CPU-imbere imbere imbere yimiyoboro ihindura amajwi, kuririmba synthesizers, hamwe nubushakashatsi bwibanze, hamwe nibiranga F0-byongeweho-byerekana-ibahasha iracyagaburira moderi nyinshi zidasanzwe. Tegereza sisitemu ya Hybrid aho isi-yuburyo busobanurwa ibipimo biyobora decodeur ya neural, biha abayiremye kugenzura neza ikibuga na timbre badatanze realism.
Gushyira mu bikorwa Isi
Ibikoresho byo guhindura amajwi bihindura ijwi ryumuvugizi na timbre mugihe imvugo yumvikana
Kuririmba amajwi ya synthesizers (nka ecosystem ya UTAU / NNSVS) isubiramo inyandiko mubibuga bishya
Sisitemu yimyandikire-y-imvugo itanga F0, spekure, na aperiodicity imigezi mbere yo gutora
Vuga ubushakashatsi bwibanze bwo guhinduranya ikibanza, kurambura igihe, no guhindura inyandiko utabanje kwitoza
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 Source-Filter Vocoding and WORLD 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
Demucs Umuziki Inkomoko Gutandukana
Ibibazo bikunze kubazwa
What is Source-Filter Vocoding and WORLD?
Indangururamajwi nigikoresho gitandukanya imvugo mubice byubaka kandi ikongera ikubaka. Inkomoko-muyunguruzi yerekana hamwe na IJWI ryijwi ryisi nuburyo bwa kera butera imbaraga inyandiko-ku-jambo no guhindura amajwi utandukanya ibyo imigozi yawe yijwi ikora nibyo umunwa wawe umeze.
Mu nkomoko-muyunguruzi yerekana imvugo, 'akayunguruzo' kagereranya iki?
Akayunguruzo ni inzira yijwi ryumvikana - imiterere yumuhogo, umunwa, nizuru bisiga amajwi. Inkomoko nijwi-umugozi buzz cyangwa urusaku rwinshi.
Nibihe bipimo bitatu amajwi y'isi akuramo imvugo?
ISI ibora imvugo muburyo bwibanze (F0), ibahasha yerekana (timbre / filter), hamwe na aperiodicity (urusaku nuburinganire bwijwi).
Algorithm YISI yitwa nde kugirango ugereranye ibahasha yerekana?
CheapTrick igereranya ibahasha itunganijwe neza kandi ikomeye ku makosa mato mu kigereranyo.
Ni ukubera iki gutandukanya ikibanza n'ibahasha ya spekral ari ingirakamaro?
Kuberako ikibanza (F0) na timbre (ibahasha yerekana) ari imigezi yigenga, urashobora kuzamura cyangwa kumanura hasi mugihe urinda indangamuntu.
Nigute ISI igereranya naba majwi bavuza amajwi nka HiFi-GAN?
ISI ni ikimenyetso-gitunganya amajwi: byihuse, birasobanurwa, hamwe na CPU. Indangururamajwi zisanzwe zigera kuri kamere karemano ariko ziremereye.