MusicLM: Hierarchical Semantic and Acoustic Tokens
MusicLM shine Google's samfurin rubutu-zuwa- kiɗa wanda ke haifar da dogon tsari mai jiwuwa ta hanyar matsayi na alamomin ma'ana don tsarin kiɗan da alamun sauti don cikakkun sauti.
Zurfafa nutsewa
An sanar da shi ta Google Bincike a farkon 2023, MusicLM yana tsara tsararrun kiɗan azaman tsinkaya jeri na alamun sauti mai ma'ana, kamar samfurin harshe yana tsinkayar kalmomi. Yana amfani da matsayi na wakilci: alamomin ma'ana (daga samfurin da ake kira w2v-BERT) yana ɗaukar tsari mai girma kamar waƙa da kari akan dogon lokaci, yayin da alamun sauti (daga SoundStream neural codec) suna ɗaukar cikakkun bayanai kamar timbre da rubutu. Mataki na farko yana haifar da alamomin ma'ana daga saƙon rubutu, sa'an nan kuma matakai na gaba suna cika cikakkun bayanan sauti da aka tsara akan waɗannan tarukan. Tsarin rubutu ya fito daga MuLM/MuLan, haɗin gwiwar rubutun kiɗa-rubutu waɗanda aka horar da su don haka kwatanci da ƙasa mai jiwuwa a sarari ɗaya. Wannan tsarin da aka tsara yana ba wa MusicLM damar kasancewa daidai da kida fiye da ɗigon ruwa bayan ƴan daƙiƙa.
Fahimtar Fasaha
Makullin ra'ayi shine ɓata tsari daga rubutu a cikin tsarin alama. Alamun ma'anar ma'anar ƙaƙƙarfan ƙaƙƙarfan alamomi ba su da yawa kuma suna canzawa sannu-sannu, don haka Transformer zai iya yin ƙira na dogon lokaci ba tare da babban tsayin jeri ba. Alamu na Acoustic suna da yawa kuma suna da girma, amma kawai suna buƙatar yin annabta sharadi akan ƙayyadaddun ƙayyadaddun ƙayyadaddun ma'anar, yin kowane mataki ana iya gani. Ƙididdigar saura vector na SoundStream yana samar da lambobi masu ƙira waɗanda na'urar ƙira ta ƙarshe ta juya baya zuwa 24 kHz waveforms.
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 MusicLM: Matsayin Semantic da Acoustic Token
Hanyar alama ta matsayi na MusicLM ta zama samfuri don tsarin baya kamar MusicGen da kayan aikin kiɗa na kasuwanci. Yi tsammanin sanyaya waƙa mai ƙarfi (hum a tune, samun cikakken tsari), tsayayyen waƙoƙi mai tsayi tare da ayoyi da mawaƙa, da ingantaccen iko akan kayan kida da maɓalli. Batutuwan ƙayayuwa sune na doka da ɗabi'a: ba da lasisin bayanan horo, izinin zane-zane, da alamar ruwa da aka samar da sauti ta yadda za'a iya bambanta shi da kiɗan da ɗan adam ke yi yanzu shine tsakiyar turawa.
Aiwatar da Gaskiyar Duniya
Juyar da rubutaccen bayanin fage zuwa fim ko makin tirela, misali. 'Epic Orchestral gini tare da mawaƙa'
Ƙirƙirar kiɗan baya da aka sharadi akan taken hoto ko ma bayanin zanen kayan aikin fasaha
Ƙaddamar da ɗan gajeren waƙa ko busa waƙa zuwa cikakken tsari na kayan aiki
Samar da nau'ikan waƙoƙi-kiɗa daban-daban a lokaci daban-daban da yanayi don talla da masu ƙirƙirar abun ciki
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 MusicLM: Hierarchical Semantic and Acoustic Tokens 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
Ƙarfin Kiɗa na Alama
Tambayoyin da ake yawan yi
What is MusicLM: Hierarchical Semantic and Acoustic Tokens?
MusicLM shine Google's samfurin rubutu-zuwa- kiɗa wanda ke haifar da dogon tsari mai jiwuwa ta hanyar matsayi na alamomin ma'ana don tsarin kiɗan da alamun sauti don cikakkun sauti.
Ta yaya MusicLM ke kula da aikin ƙirƙirar kiɗa?
MusicLM yana tsara ƙarni na kiɗa azaman tsinkayar juzu'i akan sahihan abubuwan sauti, kama da yadda ƙirar harshe ke annabta kalmomi.
Menene rawar alamomin ma'ana a cikin MusicLM?
Alamu na Semantic (daga w2v-BERT) suna ɗaukar tsattsauran ra'ayi, tsari mai saurin canzawa kamar waƙa da kari na tsawon tsayi.
Wane bangare ne ke ba da cikakkun bayanai na sauti kamar timbre da rubutu?
Alamu na Acoustic sun fito daga SoundStream neural codec kuma suna ɓoye cikakkun bayanai kamar su timbre da rubutu.
Ta yaya MusicLM ke haɗa saƙon rubutu zuwa sautin kiɗa?
An horar da MuLan ta yadda bayanin rubutu da daidaita taswirar odiyo zuwa wuraren da ke kusa a cikin sararin haɗawa da aka raba, yana ba da damar daidaita rubutu.
Me yasa tsararru, tsararrun ƙira ke taimakawa tare da tsari mai tsayi?
Alamu na ma'ana suna canzawa sannu a hankali, don haka Transformer zai iya yin tsari na dogon lokaci da nagarta sosai kafin a cika cikakkun bayanan sauti.