Vocoder ya DiffWave
DiffWave ni ikwirakwizwa rishingiye ku majwi asanisha amajwi mu buryo bwitondewe yerekana urusaku rudasanzwe mu muhengeri, rushingiye kuri mel-spectrogram.
Incamake
It brought diffusion models to high-fidelity speech, rivaling GANs and WaveNet without adversarial training.
Kwibira cyane
DiffWave, yatangijwe na Kong n'abandi. muri 2020, ikoresha icyerekezo cyerekana ikwirakwizwa ryicyitegererezo kumajwi mbisi. Mugihe cy'amahugurwa byongera buhoro buhoro urusaku rwa Gaussiya kumurongo usukuye hejuru yintambwe nyinshi, hanyuma wiga umuyoboro wo guhanura no gukuraho urwo rusaku kuri buri ntambwe. Mugihe cyibisekuru bitangirira ku rusaku rwiza kandi bigakora inzira ihindagurika, itondekanye kuri mel-spectrogram, kugirango igarure imvugo isukuye. Uruti rw'umugongo ni umuyoboro udasanzwe, wagutse-uhuza ibitekerezo bisa na WaveNet ariko ugahanura urusaku aho kuba ingero. DiffWave ihuza amajwi akomeye mu bwiza kandi irakomeye, ndetse itanga imvugo yumvikana itagabanijwe hamwe nibisubizo bihamye kubavuga. Igicuruzwa nyamukuru cyihuta ni umuvuduko: naive sampling ikenera intambwe ibihumbi kugeza ku bihumbi, nubwo gahunda yihuse igabanya ibi kugeza kuri bitandatu.
Ubushishozi
DiffWave yiga igipimo cyo gukwirakwiza amakuru mu buryo butaziguye mu guhugura urusobe rwo guhanura urusaku rwongewe ku ntambwe yo gukwirakwiza, ukoresheje intego yoroshye ya L2. Gutoranya bihindura ingengabihe y'urusaku, kandi umubare wintambwe ucuruza ubuziranenge bwihuta; abashakashatsi basanze bitondewe gahunda ngufi yintambwe zigera kuri esheshatu zirinda ubudahemuka bwinshi, bihindura inzira yintambwe igihumbi mubintu byegereye ibikorwa bifatika.
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 ka DiffWave Diffusion Vocoder
DiffWave yatangije amajwi ya diffuzione hamwe nabasimbuye byihuse nka PriorGrad na FastDiff igabanya intambwe zibarwa. Umwanya urimo guhuriza hamwe kubuhanga no guhora-bigezweho bigamije gutoranya intambwe imwe yo gukwirakwiza, kuziba icyuho cyihuta hamwe nijwi rya GAN mugihe gikomeza imyitozo ihamye kandi ikomeye. Tegereza ibitekerezo byo gukwirakwiza gukwirakwira cyane muri muzika, codecs ya neural, hamwe na majwi yisi yose aho uburyo bwo gukwirakwiza ibintu bifite akamaro.
Gushyira mu bikorwa Isi
Kwizerwa cyane neural text-to-imvugo inyuma birangira birinda imyitozo ya GAN idahindagurika
Imvugo itagabanijwe kubyara amakuru yo kongera amakuru nubushakashatsi bwamajwi
Speaker-robust ijwi synthesis aho icyitegererezo kimwe gikora amajwi menshi buri gihe
Ikizamini cyihuse-cyitegererezo cyo gukwirakwiza ubushakashatsi, ukoresheje ingengabihe ngufi y'amajwi kumajwi-nyayo
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 DiffWave Diffusion 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
Ijwi rihamye Gutandukana
Ibibazo bikunze kubazwa
What is DiffWave Diffusion Vocoder?
DiffWave ni ikwirakwizwa rishingiye ku majwi asanisha amajwi mu buryo bwitondewe yerekana urusaku rudasanzwe mu muhengeri, rushingiye kuri mel-spectrogram. Byazanye moderi yo gukwirakwiza imvugo yizerwa cyane, irwanya GAN na WaveNet nta mahugurwa yo guhangana.
Nigute DiffWave itanga amajwi mugihe cyo gufata umwanzuro?
DiffWave itangirira ku rusaku rwa Gaussiya kandi ikora inzira yo gukwirakwiza, ikagaragaza inshuro nyinshi mugihe itunganijwe kuri mel-spectrogram, kugirango itange umurongo.
Niki umuyoboro wa DiffWave wiga guhanura mugihe cy'amahugurwa?
DiffWave itoza umuyoboro wo guhanura urusaku rwa Gaussiya rwongewe ku ntambwe yatoranijwe yo gukwirakwiza, ukoresheje igihombo kiremereye L2.
Ni ubuhe butumwa bukomeye bwa DiffWave ugereranije n'abavuga GAN?
Naive diffusion sampling ikenera intambwe nyinshi zinyuranye; nubwo gahunda yihuse ifasha, inzira itera nigiciro nyamukuru cyigiciro.
Ni izihe nyungu DiffWave ifite kurusha abavuga GAN mumahugurwa?
Moderi ya Diffusion yatojwe hamwe nintego ihamye yo gusubira inyuma, ikuraho ihungabana imyitozo ya GAN ishobora guhangana.
Ni ubuhe bwubatsi umuyoboro wa DiffWave urusaku-ruhanura rusa?
DiffWave ikoresha umugongo utari autoregressive wumugongo wagutse wagutse usa na WaveNet, ariko irahanura urusaku aho gukoresha amajwi.