MelGAN Generative Vocoder
MelGAN bụ ụda olu dabere na GAN na-agbanwe agbanwe nke na-atụgharị mel-spectrograms ka ọ bụrụ ụdị ebili mmiri dị egwu n'otu ngafe na-aga n'ihu ngwa ngwa.
Nchịkọta
It mattered because it proved high-quality, non-autoregressive speech synthesis could run hundreds of times faster than real time on a GPU.
Ime miri emi
MelGAN, nke Kumar et al. na 2019, na-ewepụta ọdịyo na-enweghị ngwa ngwa-site-sample loop nke WaveNet na-eji. Ihe na-emepụta ihe na-emepụta ya bụ nchịkọta ntụgharị ntụgharị nke na-egosipụta mel-spectrogram (nke a na-ejikarị eriri ugboro 80) ruo ọnụego nlele ọdịyo, yana ihe mgbochi ndị fọdụrụ na-eji mgbagwoju anya gbasaa iji gbasaa ogige nnabata. Isi ihe ọhụrụ ewepụtara bụ ọzụzụ na ọtụtụ ịkpa oke na-arụ ọrụ n'ụgwọ ọnụnụ dị iche iche (ụdị ebili mmiri mbụ gbakwunyere ụdị agbadata), nke ọ bụla na-ele anya na windo ndị nwere oke. Ọnwụ dakọtara na-atụnyere mmemme ịkpa oke n'etiti ọdịyo adịgboroja na nke adịgboroja, na-eme ka ọzụzụ GAN guzosie ike. Ụdị ahụ dị ntakịrị site na ụkpụrụ neural-audio ma na-agba ọsọ ọsọ karịa ozugbo ọbụna na CPU, na-eme ka ọ dị irè maka ederede-gaa na-okwu agbakwunyere na ngwaọrụ.
Nghọta nka nka
MelGAN's multi-scale discriminator na-eji netwọk atọ yiri ya na-elele ọdịyo n'uju, ọkara na nkeji nkeji, nhazi ọ bụla na-esetịpụ n'ogo dị iche iche. N'ụzọ dị mkpa, MelGAN na-adabere na mfu dakọtara (L1 dị n'etiti maapụ atụmatụ ịkpa ókè nke ezigbo vs. emepụtara ọdịyo) karịa mfu nrụgharị spectrogram doro anya, nke na-agba ndị na-emepụta ọkụ ume ka ọ kwekọọ na oyi akwa ọnụ ọgụgụ ọdịyo n'ezie.
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 MelGAN Generative Vocoder
MelGAN mụrụ ezinụlọ nke GAN vocoders. Ndị ga-anọchi ya, HiFi-GAN na UnivNet, debere ụzọ ọsọ ọsọ na-abụghị nke onwe mana agbakwunyere ọtụtụ oge na oke mkpebi maka oke ugboro dị elu dị ọcha. Ihe owuwu a na-ebi na ngwaọrụ na nkwanye TTS ebe latency na ihe nlereanya nha, yana echiche ịkpa oke ya na-aga n'ihu na-emetụta codecs neural na sistemụ ọgbọ egwu ebe ọzụzụ mmegide na-eme ka mma nghọta.
Mmejuputa n'ezie n'ụwa
Ederede-gaa-okwu na ngwaọrụ n'ime ndị enyemaka mkpanaka ebe obere ụda olu na-agba ọsọ na-ezere njem okirikiri igwe ojii
Ọkpụkpọ ntụgharị ụda ozugbo na-agbanwe mel-spectrogram nke ọkà okwu ka ọ bụrụ olu ebumnuche
Ngwa egwuregwu na ihe ngosi na-emepụta mkparịta ụka agwa sitere na spectrograms ewepụtara nwere obere latency
Usoro nyocha maka GAN ọdịyo, ebe a na-ejigharị mfu dakọtara MelGAN maka egwu na ọgbọ mmetụta ụda.
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 MelGAN Generative 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
Vocoder DiffWave Diffusion
Ajụjụ a na-ajụkarị
What is MelGAN Generative Vocoder?
MelGAN bụ ụda olu dabere na GAN na-agbanwe agbanwe nke na-atụgharị mel-spectrograms ka ọ bụrụ ụdị ebili mmiri dị egwu n'otu ngafe na-aga n'ihu ngwa ngwa. Ọ dị mkpa n'ihi na ọ gosipụtara ịdị elu dị elu, nchịkọta okwu na-enweghị isi nwere ike ịgba ọsọ ọtụtụ narị ugboro ngwa ngwa karịa ozugbo na GPU.
Kedu ndenye MelGAN na-atụgharị ka ọ bụrụ ụdị ebili mmiri dị egwu?
MelGAN bụ vocoder: ọ na-ewe ihe ngosi njiri mara ụda olu, mel-spectrogram, ma na-ahazi usoro ikuku kwekọrọ.
Kedu ihe kpatara MelGAN ji dị ngwa karịa WaveNet na ntinye aka?
WaveNet na-ewepụta ihe nlele otu n'otu oge na-akpaghị aka; Igwe ọkụ na-eme mgbanwe nke MelGAN na-ewepụta ụdị ebili mmiri niile n'otu ngafe na-aga n'ihu.
Kedu ụdị ịkpa oke dị iche iche MelGAN webatara?
MelGAN na-eji ọtụtụ ịkpa oke yiri nke ahụ na-enyocha ụdị ebili mmiri mbụ na ụdị ewedatara iji weghara usoro n'ọtụtụ dị iche iche.
Kedu ọnwụ na-enyere aka kwụsie ike ọzụzụ mmegide nke MelGAN?
Ọnwụ dakọtara na njirimara na-ebelata ohere L1 dị n'etiti maapụ njirimara ịkpa oke maka ụda olu dị adị na nke ewepụtara, na-eduzi generator na ọzụzụ ọzụzụ.
Kedu ka onye na-emepụta MelGAN si abawanye ubi nnabata ya?
Ihe mgbochi ndị fọdụrụ na mkpọmkpọ ebe gbasasịrị agbasa na-agbasawanye ogige nnabata nke ọma, na-ekwe ka onye na-emepụta ọkụ na-emepụta ihe nhazi oge dị ogologo.