Vocoder DiffWave Diffusion
DiffWave bụ vocoder gbadoro ụkwụ na-agbasa nke na-emepụta ụda site n'ịkwado mkpọtụ na-enweghị usoro ka ọ bụrụ usoro ebili mmiri, dabara na mel-spectrogram.
Nchịkọta
It brought diffusion models to high-fidelity speech, rivaling GANs and WaveNet without adversarial training.
Ime miri emi
DiffWave, nke Kong et al webatara. n'afọ 2020, na-etinye usoro ihe nlere anya nke puru omume na-emebi ụda olu. N'oge ọzụzụ ọ ji nwayọọ nwayọọ na-agbakwụnye mkpọtụ Gaussian na ebili mmiri dị ọcha n'ọtụtụ usoro, wee mụta netwọk iji buo amụma ma wepụ mkpọtụ ahụ na nzọụkwụ ọ bụla. N'oge ọgbọ ọ na-amalite site na ụzụ dị ọcha ma na-eme usoro ntụgharị ahụ, na-agbanye na mel-spectrogram, iji nwetaghachi okwu dị ọcha. Ọkpụkpụ azụ bụ netwọọkụ na-abụghị nke autoregressive, gbasaa-convolution dị ka WaveNet mana na-ebu amụma mkpọtụ karịa ihe nlele. DiffWave dakọtara vocoders siri ike n'ịdị mma yana ọ siri ike, ọbụlagodi na-ewepụta okwu ezi uche na-adịghị na ya yana nsonaazụ na-agbanwe agbanwe n'ofe ndị na-ekwu okwu. Isi ahịa-apụ bụ ọsọ: nlele efu chọrọ ọtụtụ puku nzọụkwụ ruo puku kwuru puku, ọ bụ ezie na usoro nhazi ngwa ngwa belata nke a ka ole na ole dị isii.
Nghọta nka nka
DiffWave na-amụta gradient nke nkesa data n'ezoghị ọnụ site n'ịzụ netwọk ka ọ buru amụma mkpọtụ agbakwunyere na usoro mgbasa ozi na-enweghị usoro, na-eji ebumnuche L2 dị arọ dị mfe. Nlereanya na-atụgharị usoro mkpọtụ a kapịrị ọnụ, yana ọnụ ọgụgụ nke nzọụkwụ na-azụta àgwà maka ọsọ; ndị nchọpụta chọpụtara usoro dị mkpirikpi ahọpụtara nke ọma nke ihe dị ka nzọụkwụ isii na-echekwa ntụkwasị obi kachasị, na-atụgharị usoro nzọụkwụ otu puku ka ọ bụrụ ihe dị nso na bara uru.
Mmetụta atụmatụ
Nweta na iru
Ọ na-eme ka nnweta ya dịkwuo mma site na ndegharị, ịkọ akụkọ, na ntụgharị olu.
Ọnụ ego na mmefu ego
Ndị otu mgbasa ozi nwere ike ibubata ọdịyo a na-egbu maramara ngwa ngwa site na iji obere mmefu ego.
Ọsọ na ọnụ ọgụgụ
Sistemụ na-eche ihu ndị ahịa nwere ike hazie mkparịta ụka n'ọtụtụ buru ibu.
Ọdịnihu nke DiffWave Diffusion Vocoder
DiffWave wepụrụ vocoders mgbasa ozi yana ndị ga-anọchi ngwa ngwa dị ka PriorGrad na FastDiff nke slash nzọụkwụ dị mkpa. Ugbo a na-agbakọta na distillation na usoro nkwụsi ike-ihe nlereanya nke ebumnuche maka nlele mgbasa ozi otu nzọụkwụ, na-emechi oghere ọsọ ya na ndị vocoders GAN ma na-edobe ọzụzụ na ịdị ike nke mgbasa. Na-atụ anya ka echiche mgbasa ozi gbasaa n'ihu n'ime egwu, codecs neural, na ọgbọ ọdịyo zuru ụwa ọnụ ebe mkpuchi ọnọdụ dị mkpa.
Mmejuputa n'ezie n'ụwa
Nkwụsị azụ nke ederede-gaa-okwu akwara dị elu nke na-ezere ọzụzụ GAN anaghị akwụsi ike
Ọgbọ okwu na-enweghị atụ maka mmụba data na nyocha ọdịyo
Nchịkọta olu siri ike na-ekwu okwu ebe otu ụdị na-ejikwa ọtụtụ olu mgbe niile
Ebe a nwalere maka nyocha mgbasa ozi ngwa ngwa, na-etinye usoro mkpọtụ obere oge na ọdịyo oge
Ihe ize ndụ & okporo ụzọ nche
Iji olu eme ihe na ihe egwu mpụta ga-abawanye mgbe nkwenye na-efu.
Izi ezi nwere ike ịdaba n'ofe ụda olu, olumba ma ọ bụ gburugburu mkpọtụ.
Enwere ike imehie ọdịyo sịntetik dị ka ezigbo okwu na-enweghị akara doro anya.
Map mmejuputa
Nweta nkwenye doro anya maka ijide olu, imechi, na ijigharị.
Nwale ogo n'ofe ndị na-ekwu okwu dị iche iche yana ọnọdụ ndabere.
Kọwaa mgbe mmadụ ga-enyocha ma ọ bụ kwado nsonye.
Deba aha ọdịyo sịntetik ma debe ndekọ ihe ndekọ maka ịza ajụjụ.
Nọgide na-eme nchọpụta
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
Ntuziaka na-esote
Mgbasa Ọdịda Ọdịda Stable
Ajụjụ a na-ajụkarị
What is DiffWave Diffusion Vocoder?
DiffWave bụ vocoder gbadoro ụkwụ na-agbasa nke na-emepụta ụda site n'ịkwado mkpọtụ na-enweghị usoro ka ọ bụrụ usoro ebili mmiri, dabara na mel-spectrogram. Ọ wetara ụdị mgbasa ozi na okwu ntụkwasị obi dị elu, na-emegide GAN na WaveNet na-enweghị ọzụzụ mmegide.
Kedu ka DiffWave si ewepụta ọdịyo n'oge ntinye?
DiffWave na-amalite site na mkpọtụ Gaussian ma na-eme usoro mgbasa ozi na-agbanwe agbanwe, na-akatọ ugboro ugboro mgbe ọ na-etinye ya na mel-spectrogram, iji mepụta ụdị waveform.
Kedu ihe netwọk DiffWave na-amụta n'ezie ịkọ amụma n'oge ọzụzụ?
DiffWave na-azụ netwọk iji buo mkpọtụ Gaussian agbakwunyere na usoro mgbasa ozi ahọpụtara na-enweghị usoro, na-eji mfu L2 dị arọ.
Kedu ihe bụ isi ihe ndọghachi azụ nke DiffWave ma e jiri ya tụnyere GAN vocoders?
Nlele mgbasa ozi na-enweghị isi chọrọ ọtụtụ usoro mgbagha; ọ bụ ezie na nhazi oge ngwa ngwa na-enyere aka, usoro nkwuputa bụ isi ọnụ ahịa ọsọ ọsọ.
Kedu uru DiffWave nwere karịa GAN vocoders na ọzụzụ?
A zụrụ ụdị mgbasa ozi site n'ebumnobi ụdị mgbagha kwụsiri ike, na-ewepụ enweghị ntụkwasị obi nke ọzụzụ GAN onye mmegide nwere ike ịta ahụhụ.
Kedu ihe owuwu ka netwọk amụma mkpọtụ DiffWave yiri?
DiffWave na-eji ọkpụkpụ azụ na-abụghị autoregressive nke mgbagwoju anya dị ka WaveNet, mana ọ na-ebu amụma mkpọtụ karịa nlele ọdịyo.