Haiper: Ukuqhubela phambili kwe-DeepMind kunye neTikTok kwiTeksti ukuya kuGuqulo lweVidiyo

Ukhuphiswano kuphuhliso lwetekhnoloji ingqiqo kwi umbhalo ukuya kwinguqulelo yevidiyo iyafikelela amanqanaba amatsha.  Haiper, inyathelo elenziwe ngamalungu angaphambili e Google DeepMind, TikTok, kunye namaziko ophando oluphambili lwezemfundo, iqalise imodeli yayo entsha ye-eponymous yokuguqula iitekisi zibe yividiyo. Olu kukhululwa lweza kamsinya emva koko I-OpenAI ingazisa Sora, ekwazi ukuvelisa iividiyo ezinomzuzu omnye wenkcazo ephezulu kwiinkcazo zombhalo ezongeziweyo.

Haiper igqame ngokuba a ubukrelekrele bokwenziwa phambili, esekwe kwiimbono ezisisiseko kwaye eyilelwe ukuphawula isiqalo sentshukumo eya IArtificial General Intelligence (AGI), ephawulwa ngobukrelekrele obuthelekiswa nobobantu namandla okufunda ngokuzimeleyo.

I-Haiper, imodeli yokuguqula umbhalo-kwividiyo eyenziwe nguGoogle DeepMind kunye neTikTok

Kwipanorama yangoku kukubhadla okungeyonyani, Haiper Ivezwa njengemodeli yoguquko ethembisa ukutshintsha imithetho yomdlalo ekudalweni komxholo wevidiyo, ukuvumela abasebenzisi balo naliphi na inqanaba lobugcisa. yenza iividiyo umgangatho ophezulu ngokulula.

Imvelaphi kunye noPhuhliso lwe-Haiper

Isiqhamo sentsebenziswano phakathi kweengcali ezivela Google DeepMind y TikTok, UHaiper uvela njengeqonga elikwaziyo ukunikela ukusuka kwi Ukuguqulwa kombhalo kwividiyo ukumisa oopopayi bemifanekiso kunye nokupeyintwa kwakhona kwevidiyo. Ukuqaliswa kwayo kufuna ukukhuphisana neziphakamiso zamva nje ezifana nemodeli ye-OpenAI ye-Sora, nangona i-Haiper ihlukaniswa ngokugxila ekufikelelekeni kunye nokulula kokusetyenziswa.

Umxholo okhethekileyo- Cofa Apha  I-WhatsApp ibuyisela uNxibelelwano kunye noNxibelelwano olutsha lweNgxoxo yeNgxoxo

Ngubani osemva kweHaiper

Iingqondo ezisemva kweli nyathelo, uYishu Miao kunye noZiyu Wang, Dibanisa amava abo kwi-TikTok kunye ne-Google DeepMind enemvelaphi eqinileyo yokufunda ngomatshini kwiYunivesithi yaseOxford. Ukusukela ekudalweni kwayo eLondon, uHaiper akayekanga ukuvela, efuna ukwenza idemokhrasi ukuyilwa kweevidiyo ezikumgangatho ophezulu.

Yintoni eyenziwa nguHaiper

I-Haiper ibalasele ngokukwazi kwayo guqula okubhaliweyo kumxholo obonwayo, ibonelela ngeempawu ezinjengokwenza iividiyo ezisuka kwisicatshulwa, ukuhambisa imifanekiso eshukumayo kunye nokuhlelwa kwevidiyo okuphezulu. Abo banomdla banokufikelela kwiqonga layo le-intanethi, babhalise usebenzisa i-imeyile yabo kwaye baqale ukuvelisa iividiyo akukho ndleko abanye, ukubhala nje isicatshulwa esifunekayo. Okwangoku, iqonga likuvumela ukuba wenze iividiyo ezinenkcazo ephezulu ukuya kuthi ga kwimizuzwana emibini, kwaye iividiyo ezikumgangatho osezantsi zinokuhlala ukuya kuthi ga kwimizuzwana emine.

Iimpawu zeHaiper kunye nezicelo

Nge-Haiper, abasebenzisi banokulinga ngezixhobo eziphambili ukuze yenza iividiyo ezimfutshane zeHD, ukusuka kwimiyalelo ebhaliweyo elula. Nangona ngoku kunqunyelwe kwiziqeshana ukuya kuthi ga kwimizuzwana emine, iqela elingasemva kweHaiper lisebenza ekwandiseni amandla alo kunye nokufikelela.

 

Iinzuzo kunye neeNdleko zeHaiper

Omnye wemida ephambili yeHaiper ilele kwi Ubufutshane beevidiyo ezenziweyo, nangona inkampani ithi isebenza ukwandisa ixesha lezi kliphu. Okwangoku, isixhobo sinikezelwa simahla, efuna ukukhuthaza ukudalwa koluntu olusebenzayo malunga neteknoloji yalo. UMiao, omnye wabaseki, ukhankanye ukuba kusengaphambi kwexesha ukuqwalasela imodeli yobhaliso kwinkonzo yokuvelisa ividiyo.

Isicwangciso sangoku sikaHaiper sijolise ekuphuhliseni a webhsayithi umsebenzisi ugxile, ngenjongo yokudibanisa inkqubo yokudala ividiyo ehlangabezana neemfuno zabasebenzisi abahlukeneyo. Nangona kungekho nkcukacha zicacileyo zemodeli esetyenziswayo ekwabelwana ngazo, inkampani sele iqalile ukuvavanya kunye nabaphuhlisi ngokusebenzisa i-API yayo yangasese, ilindele impendulo eya kuba negalelo ekucokiseni kwayo. Kwi-horizon, ukuqaliswa kweemodeli eziphambili kunye nezikhethekileyo zicwangcisiwe.

Nangona imodeli ye-OpenAI ye-Sora ingekafumaneki kuluntu ngokubanzi, I-Haiper imema abasebenzisi ukuba bazame isixhobo sayo simahla ngewebhusayithi yabo. Nangona kunjalo, kusekho indlela ende ekufuneka ihanjwe kweli nyathelo litsha, ngakumbi xa kuqwalaselwa ubunkokeli obusekiweyo bamaziko afana ne-OpenAI kunye noGoogle kwibala le kukubhadla okungeyonyani.

Imibono yokuqala kunye nohlalutyo

Umzekelo osebenzayo watyhila isakhono se Haiper ukuvelisa ukubonakaliswa kwe-hyperrealistic kwiinkcazo zokudala, nangona kukho imida ethile kwiinkcukacha kunye nokwakheka. Olu hlalutyo lubonisa amandla amakhulu kunye nemimandla yokuphucula iqonga.

Ikamva leHaiper

Ngaphandle kokungeniswa kwayo mva nje, i-Haiper sele icwangcisa ukwandisa ukusebenza kwayo kwaye ivule i-API yayo kubaphuhlisi, ukuze ifumane ingxelo kwaye igqibelele imodeli yayo. Ngokungafaniyo ne-OpenAI's Sora, i-Haiper ngoku iyafumaneka kuluntu ukuba ivavanye isixhobo sayo simahla.

UHaiper umele ukuqhubela phambili okuchulumancisayo kwinkalo yobukrelekrele bokwenziwa esetyenziswa ekudalweni komxholo wedijithali. Ukuzibophelela kwabasunguli bayo ekuveliseni izinto ezintsha kunye nokufikeleleka kuqikelela ikamva elithembisayo lesi sixhobo.