Vocodeur de diffusion d'onde diffusion
DiffWave vocoder la bu sukkandiko ci diffusion, muy synthesize audio ci di dindi bruit yu bari yi ci forme onde, buñu sukkandikoo ci mel-spectrogram.
Résumé
It brought diffusion models to high-fidelity speech, rivaling GANs and WaveNet without adversarial training.
Plongeur bu xóot
DiffWave, Kong ak ñeneen ñoo ko dugal. ci 2020, dafay jëfandikoo kaadar modelu diffusion buy dindi bruit ci audio bu ñor. Bu ñuy tàggat, dafay yokk ndànk-ndànk bruit Gaussian ci forme vague bu sell ci jéego yu bari, ginaaw ga mu jàng reso buy seetlu ak dindi bruit boobu ci jéego bu nekk. Ci jamonoy generation dafay tàmbali ci bruit bu sell ba noppi def ci anam wu wuute, ci mel-spectrogram, ngir am kàddu yu sell. Yaxu ndigg la bu amul autoregresif, reso bu yaatu bu nuru WaveNet waaye di wax luy waaja am ci bruit bi moo gën misaal yi. DiffWave méngoo na ak vocoder yu am doole ci kalite, te dafa am doole lool, ba sax génne kàddu yu jaar yoon te amul benn waruwaay, ak njariñ yu méngoo ci waxkat yépp. Li gëna am solo mooy gaawaay bi: sampling naïf mingi soxla ay fukki-fukki jéego wala ay junni jéego, waaye oraaru gaaw yi dañuy dagg lii ci lu néew ci jiroom benn jéego.
Gis-gis xarala
DiffWave dafay jàng gradient bi ci séddaleb done ci anam wu nëbbu ci tàggat ab reso ngir xam luy bruit biy yokk ci jéego diffusion bu bariwul, jëfandikoo objectif L2 bu yomb. Sampling dafay soppi kalendriye bruit buñ tëral, ba noppi limu jéego yi dafay wecci kalite bi ak gaawaay; Gëstukat yi gis nañu ni ay jamono yu gàtt yuñ tànnee bu baax, yu am lu tollu ci 6 jéego, ñooy gëna am njub, ba noppi soppi liggéey bu am junniy jéego, mu nekk lu gëna jege jëfandikoo.
njeextalu pexe
Dugg ak yegg
Dafay gëna yombal jëfandikoo gi jaaraleko ci transkripsioŋ, nettali ak interfaasu baat.
Njëgg ak budget
Ekipu mejaa yi mën nañu yónnee audio bu leer ci anam wu gëna gaaw te seen xaalis gëna néew.
Gaawaay ak yaatuwaay
Sistem yiy jàkkarloo ak kiliyaan bi mën nañu def waxtaan ci anam wu gëna yaatu.
Ëlëgu Vocoder Diffusion DiffWave
DiffWave dafa dugal vocoder di tasaare ak ñi ko wuutu lu gëna gaaw lu melni PriorGrad ak FastDiff yuy dagg jéego yi. Doxalin bi mingi jëm ci distilaasioŋ ak pexe yu méngoo-model yiy fexe jël misaalu diffusion ci benn jéego, tëj gap bi am ci gaawaay bi ak vocoders GAN boole ci tëye diffusion bi ci tàggat ak dëgër. Xaarandil xalaati diffusion ñu gëna tasaaroo ci music, codec neuronal, ak defar audio universel fu coverage mode am solo.
Doxal ci àdduna dëgg
Fidélité neural bu kawe bind-ci-wax ginaaw bi moytu tàggat GAN bu amul dal
Defar kàddu yu amul benn sartu yokk ay done ak gëstu audio
Synthese baat bu dëgër buy wax, fu benn model mëna jëfandikoo baat yu bari ci anam wu méngoo
Testbed ngir gëstu diffusion sampling bu gaaw, jëfandikoo oraaru bruit yu gàtt ci audio ci jamono dëgg
Risk yi ak balustrade yi
Jëfandikoo baat ci anam wu jaarul yoon ak niru ak nit dafay gëna yokk sudee nanguwul.
Jaar-jaar mën na wàññeeku ci aksan yi, dialect yi wala barab yu bari xumbaay.
Audio synthetik mën nañu ko jaawale ak wax ju dëggu sudee amul etiket bu leer.
Roadmap ngir samp gi
Wutal ndigal bu leer ngir jàpp baat bi, klone ko ak jëfandikoowaat ko.
Saytu kalite ci kàddukat yu bari ak anam yu bari ci ginaaw.
Mandargal kañ la nit wara xoolaat wala nangu ay génne.
Etiketu audio synthetik te nga denc dokimaa ci fimu bawoo ngir mëna lim.
Weyal di banneexu
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
Gis bi ci topp
Diffusion audio bu nëbbu bu dëgër
Laaj yi ñuy faral di laaj
What is DiffWave Diffusion Vocoder?
DiffWave vocoder la bu sukkandiko ci diffusion, muy synthesize audio ci di dindi bruit yu bari yi ci forme onde, buñu sukkandikoo ci mel-spectrogram. Dafa indi xeetu diffusion ci wax ju dëggu, di xëccoo GANs ak WaveNet te kenn tàggatul noon.
naka DiffWave di defaree audio ci waxtu inférence?
DiffWave mingi tàmbali ci bruit Gaussian ba noppi def ci diffusion inverse, di denoiser lu bari yoon ci biir mel-spectrogram, ngir génne forme vague.
Lan la reso DiffWave di jàng xam luy waaja am ci diiru tàggat yaram?
DiffWave dafay tàggat ab reso ngir mëna wax luy waaja am ci bruit Gaussian biñ yokk ci jéego diffusion buñu tann ci anam wu mucc, di jëfandikoo L2 buñ pondéree.
Lan mooy njariñu DiffWave buñu ko méngale ak vocoder GAN yi?
Échantillonnage diffusion naïf mingi soxla jéego yu bari yu ñu delloo ginaaw; Doonte oraaru gaaw yi dañuy jàppale, liggéey bi ñuy baamtu mooy njëgu gaawaay bi gëna mag.
Lan moo gëna am solo ci DiffWave ci kaw vocoder GAN yi ci tàggat yaram?
Modèle diffusion yi dañu leen di tàggat ci anam regression bu dëgër, di moytu instabilite bi tàggat GAN buy xeex mëna am.
Ban architecture la reso buy wax luy waaja am ci DiffWave nuru?
DiffWave dafay jëfandikoo yaxu ndigg bu dul dellu ginaaw ci ay convolution yu yaatu yu niro ak WaveNet, waaye dafay wax luy waaja am ci bruit bi moo gën misaali audio yi.