MelGAN Generative Vocoder
MelGAN cikakken sauti ne na tushen GAN mai jujjuyawa wanda ke juya mel-spectrograms zuwa sifofin raƙuman sauti mai sauƙi a cikin wucewar gaba ɗaya cikin sauri.
Dubawa
It mattered because it proved high-quality, non-autoregressive speech synthesis could run hundreds of times faster than real time on a GPU.
Zurfafa nutsewa
MelGAN, Kumar et al. a cikin 2019, yana haifar da sauti ba tare da jinkirin samfurin-samfurin madauki wanda WaveNet ke amfani dashi ba. janaretansa tarin juzu'i ne wanda ya samar da misalan mel-spectrogram (yawanci madaurin mitar mitar 80) har zuwa ƙimar samfurin sauti, tare da ragowar tubalan ta amfani da faɗuwar juzu'i don faɗaɗa filin karɓa. Mabuɗin ƙirƙira shine horarwa tare da masu wariya da yawa waɗanda ke aiki a ma'auni na sauti daban-daban (tsarin ƙaƙƙarfan ƙaƙƙarfan ƙaƙƙarfan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan nau'ikan suna kallon windows masu mamayewa. Asarar da ta dace da fasalin tana kwatanta kunnawar wariya tsakanin sauti na gaske da na karya, yana tabbatar da horon GAN. Samfurin yana ƙarami ta ma'auni-audio kuma yana aiki da sauri fiye da ainihin lokacin har ma akan CPU, yana mai da shi mai amfani don rubutu-zuwa-magana na na'ura.
Fahimtar Fasaha
MelGAN's Multi-scale discriminator yana amfani da cibiyoyin sadarwa iri ɗaya guda uku suna kallon sauti gabaɗaya, rabin, da ƙuduri kwata, kowane tsari mai ɗaukar hoto a jeri daban-daban. Mahimmanci, MelGAN ya dogara da hasarar da ta dace (nisa ta L1 tsakanin taswirar fasalin nuna wariya na ainihin sauti da aka haifar) maimakon fayyace asarar sake ginawa, wanda ke ƙarfafa janareta don dacewa da ainihin ƙirar ƙididdiga na audio ta Layer.
Dabarun Tasiri
Shiga ku isa
Yana inganta samun dama ta hanyar rubutu, ba da labari, da mu'amalar murya.
Kudin da kasafin kuɗi
Ƙungiyoyin kafofin watsa labaru na iya jigilar sauti mai gogewa cikin sauri tare da ƙaramin kasafin kuɗi.
Gudu da sikelin
Tsarin fuskantar abokin ciniki na iya aiwatar da hulɗar magana a mafi girman ma'auni.
Makomar MelGAN Generative Vocoder
MelGAN ya haifi dangin GAN vocoders. Magadansa, HiFi-GAN da UnivNet, sun kiyaye hanyar da ba ta dace ba cikin sauri amma sun ƙara yawan lokuta da masu bambance-bambance masu yawa don tsaftataccen mitoci. Gine-ginen yana rayuwa a cikin na'ura da watsawa TTS inda latency da samfurin girman al'amura, da ra'ayoyinsa na nuna wariya suna ci gaba da yin tasiri ga codecs na jijiyoyi da tsarin tsara kiɗan inda horarwar adawa ke inganta ingancin fahimta.
Aiwatar da Gaskiyar Duniya
Rubutu-zuwa-magana a kan na'ura a cikin mataimakan wayar hannu inda ƙarami, mai sauti mai sauri ya guje wa tafiye-tafiyen gajimare
Bututun canza murya na lokaci-lokaci waɗanda ke canza mel-spectrogram na mai magana zuwa muryar manufa
Wasan kwaikwayo da kayan aikin raye-raye waɗanda ke haɗa maganganun halayyar daga spectrograms da aka ƙirƙira tare da ƙarancin latency
Tushen bincike don GANs mai jiwuwa, inda aka sake amfani da asarar madaidaicin fasalin MelGAN don kiɗan da tasirin sauti.
Hatsari & Tsare-tsare
Rashin amfani da murya da haɗarin kwaikwaya yana ƙaruwa lokacin da aka rasa izini.
Daidaituwa na iya faɗuwa cikin lafuzza, yaruka, ko mahalli masu hayaniya.
Ana iya kuskuren sauti na roba don ingantacciyar magana ba tare da bayyananniyar lakabi ba.
Taswirar Hanya
Sami tabbataccen izini don ɗaukar murya, cloning, da sake amfani.
Gwajin ingantattun masu magana daban-daban da yanayin baya.
Ƙayyade lokacin da dole ne ɗan adam ya duba ko ya amince da abubuwan da aka fitar.
Yi lakabin sauti na roba da kuma adana bayanan da aka tabbatar don yin lissafi.
Ci gaba da Bincike
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
Jagora na gaba
DiffWave Diffusion Vocoder
Tambayoyin da ake yawan yi
What is MelGAN Generative Vocoder?
MelGAN cikakken sauti ne na tushen GAN mai jujjuyawa wanda ke juya mel-spectrograms zuwa sifofin raƙuman sauti mai sauƙi a cikin wucewar gaba ɗaya cikin sauri. Yana da mahimmanci saboda ya tabbatar da inganci mai inganci, haɗin magana mara ƙarfi na iya tafiyar da ɗaruruwan sau da sauri fiye da ainihin lokacin akan GPU.
Wace shigarwa MelGAN ke jujjuyawa zuwa siffan sautin raƙuman sauti?
MelGAN vocoder ne: yana ɗaukar fasalin fasalin sauti, mel-spectrogram, kuma yana haɗa nau'in igiyar ruwa mai dacewa.
Me yasa MelGAN yayi sauri fiye da WaveNet akan ra'ayi?
WaveNet yana haifar da samfurori ɗaya bayan ɗaya kai tsaye; Generator mai jujjuyawar MelGAN yana samar da gabaɗayan sigar motsi a cikin layi ɗaya na gaba ɗaya.
Wane irin salo na wariya MelGAN ya gabatar?
MelGAN yana amfani da masu wariya iri ɗaya da yawa waɗanda ke yin nazarin sigar ƙaƙƙarfan ƙaƙƙarfan ƙaƙƙarfan ƙaya don ɗaukar tsari a ma'auni daban-daban.
Wace asara ce ke taimakawa daidaita horon abokan gaba na MelGAN?
Asarar da ta dace da fasalin tana rage tazarar L1 tsakanin taswirar fasalin nuna wariya don ainihin sauti da aka samar, yana jagorantar janareta da ƙarfafa horo.
Ta yaya janareta na MelGAN ke haɓaka filin karɓar sa?
Ragowar tubalan tare da faɗuwar juzu'i suna faɗaɗa filin karɓa da kyau, suna barin janareta ya ƙirƙira tsarin dogon zango.