Google tshaj tawm MusicLM, kev txawj ntse uas tsim suab paj nruag tawm ntawm cov lus koj ntaus, zoo li DALL-E 2. Nws yog hom lus tsim los ntawm Google Kev Tshawb Fawb. Tsis tas li ntawd, lawv tau tsim tshwj xeeb rau kev tsim suab paj nruag.
Thiab, nws tau raug cob qhia ntawm cov ntaub ntawv loj ntawm cov ntaub ntawv suab paj nruag thiab tuaj yeem tsim cov suab paj nruag hauv ntau hom thiab cov ntawv. Yog tias koj txaus siab rau suab paj nruag; ces koj yuav tsum xyuas seb MusicLM yuav muab dab tsi.
Nrog MusicLM koj tsim suab paj nruag hauv qee cov tswv yim thiab cov ntawv. Piv txwv li, koj tuaj yeem tsim cov piano, nruas ntaus, thiab suab paj nruag rau cov nkauj.
Tsis tas li ntawd, koj tuaj yeem kho kom zoo rau qee yam qauv lossis suav nrog cov tswv yim uas tus neeg siv tau muab. Nws yog tsim los tsim suab paj nruag uas yog harmonically thiab rhythmically cohesive. Yog li, cia peb dhia dej thiab saib seb MusicLM yog dab tsi.
Kev sim yav dhau los
MusicLM tsis yog thawj AI-tsim suab paj nruag system. Riffusion, Dance Diffusion, Google's AudioML, thiab OpenAI's Jukebox yog cov piv txwv ntawm kev sib piv. Txawm li cas los xij, cov txheej txheem ua ntej no tau txwv los ntawm kev txwv thev naus laus zis.
Tsis tas li ntawd, lawv cov ntaub ntawv tsis muaj kev cob qhia ua rau nws nyuaj rau kev sau cov suab paj nruag zoo. Txawm li cas los xij, MusicLM muaj lub peev xwm los tsim suab paj nruag nrog qib siab dua ntawm kev txawj ntse thiab kev muaj tiag.
Txheej txheem cej luam MusicLM
MusicLM kawm cov qauv thiab style ntawm suab paj nruag. Li no, nws tau txais kev cob qhia ntawm cov ntaub ntawv loj heev ntawm MIDI thiab cov ntaub ntawv suab paj nruag cim. Zoo li nws cov kev pabcuam zoo sib xws, MusicLM yog tsim los ntawm Transformer architecture.
Siv cov tswv yim rau tus kheej kom mloog zoo rau cov tswv yim tshwj xeeb, MusicLM's transformer architecture yog siv los rho tawm cov qauv thiab cov suab paj nruag los ntawm cov ntaub ntawv loj. Yog li ntawd, koj tuaj yeem tsim cov suab paj nruag harmonically thiab rhythmically cohesive.
Thiab, cov suab paj nruag no tuaj yeem ua raws li lub koom haum ntawm cov neeg siv tswv yim. Li no, koj yuav tuaj yeem tau txais cov txiaj ntsig suab paj nruag uas koj tau piav qhia tshwj xeeb rau qhov program.
Kev vam meej ntawm yav dhau los cov qauv lus, xws li GPT-2 thiab GPT-3, uas tau ua pov thawj lawv lub peev xwm los tsim kev sib raug zoo thiab kev sau ntawv zoo, tau tshoov siab rau MusicLM. MusicLM, ntawm qhov tod tes, yog thawj hom lus uas tau tsim tshwj xeeb rau cov suab paj nruag tiam.
Thiab, peb xav tias nws yuav raug suav tias yog ib qho ntawm cov qauv zoo tshaj plaws.
Nws Ua Haujlwm Li Cas?
DALL-E 2 thiab Google's MusicLM artificial txawj ntse qhia ntau yam zoo sib xws. Lub sijhawm no, txawm li cas los xij, koj cov ntawv sau tau nthuav tawm suab paj nruag ntau dua li qhov pom. Hauv qhov no, koj tuaj yeem tsim ib qho tag nrho. Tsis tas li ntawd, koj tuaj yeem tsim suab paj nruag siv ib qho twj paj nruag xwb.
Koj tuaj yeem pom ntau cov qauv kev tshawb fawb tsim los ntawm Google AI pab neeg ntawm MusicLM's Github nplooj ntawv. Txawm hais tias AI tseem nyob rau theem kev tshawb fawb thiab kev txhim kho, lub suab nws tuaj yeem ua tau yog qhov kev daws teeb meem siab. Tsis tas li ntawd, muaj cov lus qhia, xws li kev koom ua ke AI no nrog ChatGPT. Qhov kev sib koom ua ke no tuaj yeem ua rau cov suab paj nruag zoo nkauj dua thiab muaj tswv yim.
Los ntawm Humming rau Hit Melodies
MusicLM muab plaub tus qauv AI sib txawv: MuLan, AudioLM, w2v-BERT, thiab Soundstream. Txawm hais tias txhua tus qauv no muaj cov txheej txheem tshwj xeeb. Txawm li cas los xij, thaum lawv tau koom ua ke, lawv ua rau MusicLM!
Cov kws ntaus suab paj nruag thiab cov kws tshaj lij kev lag luam tau ceeb toom ntawm MusicLM lub peev xwm los hloov pauv txawm tias qhov yooj yim tshaj plaws hums thiab yws rau hauv tag nrho cov suab. Los ntawm kev sib txuas nrog ChatGPT, nws tuaj yeem tsim cov suab paj nruag tshwj xeeb.
Koj tuaj yeem mloog thiab tshawb xyuas cov suab paj nruag thiab suab tsim los ntawm MusicLM ntawm nws website. Tab sis, nco ntsoov tias tam sim no nws yog nyob rau theem kev sim. Nws yog qhov pom tseeb tias MusicLM muaj peev xwm hloov pauv kev lag luam suab paj nruag tag nrho raws li kev tsim kho tshuab.
AI-Generated Music nrog tib neeg zoo li Nuances
Txhawm rau tsim cov nkauj uas ua rau muaj kev nkag siab zoo raws li cov lus piav qhia meej, MusicLM tau kawm ntawm cov ntaub ntawv loj ntawm 280,000 teev suab paj nruag. Piv txwv li, koj tuaj yeem tsim "kev ua suab paj nruag dubstep nrog lub suab bass thaj tsam sib sib zog nqus thiab cov suab nruas nrov nrov". Los yog, koj tuaj yeem thov kom nws tsim "ib zaj nkauj uas ntxim nyiam nrog lub suab paj nruag ntxim nyiam thiab lub suab nrov." Koj txoj kev xav yog qhov txwv hauv qhov no.
Cov nkauj uas tsim tau zoo ib yam li cov neeg ua suab paj nruag. MusicLM cov qauv yog qhov xav tsis thoob. Nws yog qhov tseeb tshwj xeeb tshaj yog tias tsis muaj tib neeg koom nrog hauv cov txheej txheem sib xyaw. MusicLM tuaj yeem rov ua qhov tsis zoo xws li suab paj nruag riffs, suab paj nruag, thiab kev xav. Tsis tas li ntawd, nws ua haujlwm txawm tias muab cov ntsiab lus nyuaj thiab meej meej.
Tseem ceeb nta
Painting Caption Conditioning
Painting Caption Conditioning yog MusicLM muaj nuj nqi. Koj tuaj yeem tsim suab paj nruag raws li cov lus piav qhia lossis "caption" ntawm daim duab. Qhov no txhais tau hais tias MusicLM muaj peev xwm tsim cov suab paj nruag uas ntes cov kev xav, lub siab xav, thiab cov tswv yim qhia hauv daim duab. Qhov kev muaj peev xwm no muaj txiaj ntsig zoo rau kev ua suab paj nruag rau cov yeeb yaj kiab, video ua si, thiab txhua yam kev tshaj tawm pom.
Zaj dabneeg hom
Story Mode feature siv cov ntawv dab neeg raws li kev nkag. Li no, nws tsim cov suab paj nruag keeb kwm yav dhau. Cov neeg siv tuaj yeem siv txoj haujlwm no los tsim cov suab paj nruag rau zaj dab neeg, kev ua yeeb yaj kiab, lossis yeeb yaj kiab los ntawm kev piav qhia qhov xwm txheej lossis lub suab nrov.
Dab Neeg Hom yog ib qho cuab yeej siv rau cov kws tshaj lij xov xwm. Yog li, nws tuaj yeem tsim ntau yam ntawm cov suab paj nruag thiab cov twj paj nruag. MusicLM's Tale Mode tau txais los ua kom muaj kev xav ntawm qhov xwm txheej. Yog li, cov neeg saib tuaj yeem muaj qib ntxiv ntawm kev nkag mus rau hauv zaj dab neeg.
Tus Kws Tshaj Lij Tshaj Lij Tshaj Lij
Koj tuaj yeem kho qhov nyuaj ntawm cov nkauj tsim. Cov neeg siv tuaj yeem xaiv ntawm peb theem raws li lawv qhov kev txawj ntse. Tsis tas li, lawv tuaj yeem hais qhia qhov kev xav tau ntawm qhov nyuaj: pib, nruab nrab, lossis qib siab.
Cov yeeb yam no pab koj yog tias koj muaj kev paub suab paj nruag me ntsis thiab xav sim nrog cov nkauj tshiab. Txawm li cas los xij, yog tias koj yog ib tus kws paub txog tshuab raj, koj tuaj yeem tsim cov suab paj nruag sophisticated thiab subtle. MusicLM lub hom phiaj nrog qhov tshwj xeeb no yog xa cov kev paub siv tau rau txhua tus neeg siv.
Generation Diversity
Nrog rau Generation Diversity muaj nuj nqi, koj tuaj yeem tsim ntau lub suab nkauj ntawm tib lub tswv yim. Thiab, koj tuaj yeem muaj ntau yam ntawm cov txiaj ntsig. Qhov no txhais tau hais tias AI tuaj yeem tsim ntau hom nkauj ntawm ib zaj nkauj.
Tsis tas li ntawd, muaj lwm cov suab paj nruag lossis chord kev vam meej, thaum tseem khaws cov nkauj cov qauv thiab cov qauv. Qhov tshwj xeeb no pab AI lub suab paj nruag tsim kom muaj tswv yim ntau dua. Li no, nws ua suab paj nruag creation ntau analogous rau tib neeg songwriting.
Muaj peev xwm txwv ntawm MusicLM
Google tseem tsis tau ua MusicLM rau cov pej xeem sawv daws vim nws tseem tab tom txhim kho. Li no, koj tseem tsis tuaj yeem muab qee cov qauv ntawm cov suab paj nruag uas MusicLM tuaj yeem tsim tau. Tsis tas li ntawd, nws tseem me ntsis tsis paub txog qhov txwv MusicLM tuaj yeem muaj.
Raws li cov thev naus laus zis tseem nyob rau theem pib, nws tuaj yeem muaj qee yam kev txwv ntawm lub suab paj nruag uas tsim los lossis nws muaj peev xwm los tswj cov khoom siv tshwj xeeb.
Cov qauv tsim 'distorted zoo yog ib qho ntawm cov teeb meem tseem ceeb. Qhov no yog qhov tsim nyog los ntawm cov txheej txheem kev cob qhia siv los tsim MusicLM.
Lwm qhov tsis zoo yog tias, txawm tias MusicLM lub peev xwm los tsim cov suab. Qhov no suav nrog cov nkauj hu nkauj. Cov "lyrics" tsim los ntawm MusicLM qee zaum zoo li gibberish. Tsis tas li ntawd, lawv tuaj yeem nyuaj rau kev nkag siab. Txawm li cas los xij, MusicLM tseem tab tom txhim kho thiab cov teeb meem no tuaj yeem txhim kho.
Zaum Kawg
Thaum kawg, peb ntseeg tias cov thev naus laus zis hauv Google MusicLM yog ob qho tib si nthuav thiab ntxim nyiam. Nws yog qhov xav tsis thoob tias AI tuaj yeem ua suab paj nruag hauv ntau hom, nrog rau qib siab dua ntawm qhov tseeb. MusicLM muaj peev xwm hloov pauv kev lag luam suab paj nruag. Thiab, peb zoo siab los saib seb qhov thev naus laus zis no hloov zuj zus li cas.
Sau ntawv cia Ncua