{"id":2635,"date":"2024-06-20T12:08:24","date_gmt":"2024-06-20T12:08:24","guid":{"rendered":"https:\/\/howtogeek.blog\/lv\/?p=2635"},"modified":"2024-06-20T12:08:24","modified_gmt":"2024-06-20T12:08:24","slug":"this-new-v2a-tool-from-google-deepmind-could-be-the-last-piece-of-the-puzzle-for-ai-generated-movies-lv","status":"publish","type":"post","link":"https:\/\/howtogeek.blog\/lv\/this-new-v2a-tool-from-google-deepmind-could-be-the-last-piece-of-the-puzzle-for-ai-generated-movies-lv\/","title":{"rendered":"\u0160is jaunais Google DeepMind V2A r\u012bks var\u0113tu b\u016bt p\u0113d\u0113jais m\u012bklas gabals AI rad\u012btaj\u0101m film\u0101m"},"content":{"rendered":"

Kad tika izlaists pirmais AI \u0123ener\u0113tais video, neviens nevar\u0113ja uzmin\u0113t, ka video \u0123ener\u0113\u0161anas AI r\u012bki non\u0101ks tik t\u0101lu tik \u012bs\u0101 laika posm\u0101. Tom\u0113r \u0161odien mums ir neskait\u0101mas platformas, kas lietot\u0101jiem \u013cauj \u0123ener\u0113t augstas kvalit\u0101tes, neticami detaliz\u0113tus videoklipus, piem\u0113ram, Synthesia un Luma AI’s Dream Machine. Tom\u0113r joproj\u0101m ir da\u017ei izaicin\u0101jumi, kas ne\u013cauj \u0161iem r\u012bkiem k\u013c\u016bt par pla\u0161i izplat\u012btiem.<\/p>\n

Un liel\u0101kais, iesp\u0113jams, ir audio \u0123ener\u0113\u0161anas process. Lai gan liel\u0101k\u0101 da\u013ca video \u0123ener\u0113\u0161anas platformu var rad\u012bt labas kvalit\u0101tes video, tie liel\u0101koties ir klusi video bez audio. Pat ja ir audio, tas parasti tiek pievienots atsevi\u0161\u0137i un neatbilst lietot\u0101ja cer\u012bb\u0101m.<\/p>\n

Piem\u0113ram, ja apmekl\u0113jat Luma AI Dream Machine lapu, varat redz\u0113t da\u017eus \u013coti iespaid\u012bgus videoklipus, ta\u010du tos pavado\u0161\u0101 ska\u0146a ir diezgan visp\u0101r\u012bga un zemas kvalit\u0101tes. Ta\u010du tas var main\u012bties, izmantojot Google jauno video-audio (V2A) tehnolo\u0123iju.<\/p>\n

Tas sola nodro\u0161in\u0101t labas kvalit\u0101tes audio \u0123ener\u0113\u0161anu videoklipiem, kas noz\u012bm\u0113, ka tas beidzot \u013caus jums izveidot m\u0101ksl\u012bg\u0101 intelekta \u0123ener\u0113tas filmas ar atbilsto\u0161iem ska\u0146u celi\u0146iem un audio, p\u0101rsp\u0113jot visus AI rad\u012btos videoklipus, kas pa\u0161laik tiek ra\u017eoti.<\/p>\n

\n

AI rad\u012bts audio priek\u0161<\/span><\/p>\n<\/figcaption><\/figure>\n

https:\/\/www.youtube.com\/watch?v=VYjZlF6m3nQ<\/a><\/p>\n

Kas ir Google DeepMind video-audio izp\u0113te?<\/h2>\n

Video-to-Audio (V2A) tehnolo\u0123ija, ko izstr\u0101d\u0101jusi Google DeepMind, ir izstr\u0101d\u0101ta, lai izveidotu ska\u0146u celi\u0146us AI \u0123ener\u0113tiem videoklipiem. \u0160\u012b tehnolo\u0123ija \u013cauj \u0123ener\u0113t video un audio vienlaikus, apvienojot dabisk\u0101s valodas uzvednes ar video pikse\u013ciem, lai rad\u012btu ska\u0146as neatkar\u012bgi no videoklip\u0101 notieko\u0161aj\u0101m darb\u012bb\u0101m.<\/p>\n

\u0160o tehnolo\u0123iju var savienot p\u0101r\u012b ar AI mode\u013ciem, ko izmanto videoklipu \u0123ener\u0113\u0161anai, piem\u0113ram, Veo, un t\u0101 var pal\u012bdz\u0113t izveidot re\u0101listiskus dialogus un ska\u0146as efektus, k\u0101 ar\u012b dramatiskus rezult\u0101tus, kas atbilst videoklipam. V\u0113l svar\u012bg\u0101k ir tas, ka jaun\u0101 V2A tehnolo\u0123ija neaprobe\u017eojas tikai ar videoklipiem, kas \u0123ener\u0113ti, izmantojot AI, bet to var ar\u012b izmantot, lai \u0123ener\u0113tu ska\u0146u celi\u0146us videoklipiem, kas veidoti tradicion\u0101l\u0101 veid\u0101. T\u0101d\u0113j\u0101di varat to izmantot m\u0113m\u0101m film\u0101m, arh\u012bva materi\u0101liem un citiem.<\/p>\n

V2A tehnolo\u0123ija \u013cauj lietot\u0101jiem \u0123ener\u0113t neierobe\u017eotu skaitu videoklipu ska\u0146u celi\u0146u un pat izmantot pozit\u012bvas un negat\u012bvas uzvednes, lai vad\u012btu ska\u0146as \u0123ener\u0113\u0161anas procesu un viegli ieg\u016btu nepiecie\u0161am\u0101s ska\u0146as. Tas ar\u012b nodro\u0161ina liel\u0101ku elast\u012bbu, lai j\u016bs var\u0113tu eksperiment\u0113t ar da\u017e\u0101diem rezult\u0101tiem un atrast to, kas ir vislab\u0101kais konkr\u0113tajam videoklipam.<\/p>\n

\n

Audio paraugs ar med\u016bzu, kas puls\u0113 zem \u016bdens.<\/span> Avots: Google<\/p>\n<\/figcaption><\/figure>\n

https:\/\/www.youtube.com\/watch?v=9Q0-t8D9XFI<\/a><\/p>\n

K\u0101 darbojas V2A tehnolo\u0123ija?<\/h2>\n

Saska\u0146\u0101 ar Google teikto, uz\u0146\u0113mums eksperiment\u0113ja ar dif\u016bzijas un autoregres\u012bv\u0101m metod\u0113m un atkl\u0101ja, ka pirmais ir vispiem\u0113rot\u0101kais ska\u0146as rad\u012b\u0161anai. Tas rada \u013coti re\u0101listiskas ska\u0146as un darbojas, kod\u0113jot video saspiest\u0101 form\u0101t\u0101.<\/p>\n

P\u0113c tam dif\u016bzijas modelis tiek izmantots, lai no video atdal\u012btu nejau\u0161us trok\u0161\u0146us, pa\u013caujoties uz dabisk\u0101s valodas uzvedn\u0113m un video. Uzvednes pal\u012bdz \u0123ener\u0113t re\u0101listisku audio, kas ir ide\u0101li sinhroniz\u0113ts ar video. Tam seko audio dekod\u0113\u0161ana, p\u0113c kuras tas tiek p\u0101rveidots par audio vi\u013c\u0146u formu un apvienots ar video.<\/p>\n

Google DeepMind sniedza vair\u0101k inform\u0101cijas, lai apm\u0101c\u012btu AI, t\u0101p\u0113c lietot\u0101ji var vad\u012bt audio \u0123ener\u0113\u0161anas procesu l\u012bdz vajadz\u012bgaj\u0101m ska\u0146\u0101m un \u013cauj platformai rad\u012bt augst\u0101kas kvalit\u0101tes audio. \u0160\u0101da inform\u0101cija ietv\u0113ra run\u0101to dialogu transkriptus un detaliz\u0113tus ska\u0146u aprakstus ar AI \u0123ener\u0113t\u0101m anot\u0101cij\u0101m.<\/p>\n

Apg\u016bstot \u0161\u0101du inform\u0101ciju, V2A tehnolo\u0123ija var saist\u012bt da\u017e\u0101das vizu\u0101l\u0101s ainas ar konkr\u0113tiem audio notikumiem.<\/p>\n

\"\"
V2A tehnolo\u0123ijas darb\u012bba.<\/span> Avots: Google<\/figcaption><\/figure>\n

Kas ir pie apv\u0101r\u0161\u0146a?<\/h2>\n

DeepMind V2A tehnolo\u0123ija darbojas daudz lab\u0101k nek\u0101 citi V2A risin\u0101jumi, jo tai ne vienm\u0113r ir nepiecie\u0161ama teksta uzvedne un t\u0101 var saprast video pikse\u013cus. Ska\u0146as izvade ar\u012b nav manu\u0101li j\u0101saska\u0146o ar video. Tom\u0113r joproj\u0101m past\u0101v da\u017ei tehnolo\u0123ijas ierobe\u017eojumi, kurus Google cen\u0161as p\u0101rvar\u0113t, veicot turpm\u0101kus p\u0113t\u012bjumus.<\/p>\n

Piem\u0113ram, \u0123ener\u0113t\u0101 audio kvalit\u0101te ir atkar\u012bga no k\u0101 ievades izmantot\u0101 video kvalit\u0101tes. Ja videoklip\u0101 ir izkrop\u013cojumi vai artefakti, m\u0101ksl\u012bg\u0101 intelekta modelis tos nesaprot, jo tie nav iek\u013cauti apm\u0101c\u012bb\u0101, k\u0101 rezult\u0101t\u0101 samazin\u0101s audio kvalit\u0101te.<\/p>\n

Turkl\u0101t attiec\u012bb\u0101 uz videoklipiem, kuros ir cilv\u0113ka runa, uz\u0146\u0113mums str\u0101d\u0101, lai uzlabotu l\u016bpu sinhroniz\u0101ciju. V2A tehnolo\u0123ija m\u0113\u0123ina \u0123ener\u0113t runu, izmantojot ievades at\u0161ifr\u0113jumus, un p\u0113c tam saska\u0146ot to ar video rakstz\u012bmju l\u016bpu kust\u012bb\u0101m. Tom\u0113r, ja videoklips nav balst\u012bts uz at\u0161ifr\u0113jumiem, audio un l\u016bpu kust\u012bbas neatbilst.<\/p>\n

Ar lab\u0101k\u0101m audio \u0123ener\u0113\u0161anas iesp\u0113j\u0101m AI mode\u013ci var\u0113s \u0123ener\u0113t videoklipus, kas ne tikai izskat\u0101s iespaid\u012bgi, bet ar\u012b izklaus\u0101s lieliski. Google ar\u012b integr\u0113 savu V2A tehnolo\u0123iju ar SynthID, kas apz\u012bm\u0113 \u016bdensz\u012bmes visam saturam, kas \u0123ener\u0113ts, izmantojot AI. Tas var pal\u012bdz\u0113t nov\u0113rst t\u0101 \u013caunpr\u0101t\u012bgu izmanto\u0161anu, nodro\u0161inot piln\u012bgu dro\u0161\u012bbu.<\/p>\n

Turkl\u0101t uz\u0146\u0113mums saka, ka r\u016bp\u012bgi p\u0101rbaud\u012bs savu V2A tehnolo\u0123iju pirms t\u0101s izlai\u0161anas sabiedr\u012bbai. L\u012bdz \u0161im, \u0146emot v\u0113r\u0101 to, ko Google ir demonstr\u0113jis un sol\u012bjis n\u0101kotnei, \u0161\u012b tehnolo\u0123ija ir noz\u012bm\u012bgs sasniegums audio \u0123ener\u0113\u0161an\u0101 AI \u0123ener\u0113tiem videoklipiem.<\/p>\n","protected":false},"excerpt":{"rendered":"

Kad tika izlaists pirmais AI \u0123ener\u0113tais video, neviens nevar\u0113ja uzmin\u0113t, ka video \u0123ener\u0113\u0161anas AI r\u012bki non\u0101ks tik t\u0101lu tik \u012bs\u0101 laika posm\u0101. Tom\u0113r \u0161odien mums ir neskait\u0101mas platformas, kas lietot\u0101jiem \u013cauj \u0123ener\u0113t augstas kvalit\u0101tes, neticami detaliz\u0113tus videoklipus, piem\u0113ram, Synthesia un Luma AI’s Dream Machine. Tom\u0113r joproj\u0101m ir da\u017ei izaicin\u0101jumi, kas ne\u013cauj \u0161iem r\u012bkiem k\u013c\u016bt par […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[182,130],"class_list":["post-2635","post","type-post","status-publish","format-standard","hentry","category-how-to","tag-artificial-intelligence","tag-microsoft"],"acf":[],"_links":{"self":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2635","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/comments?post=2635"}],"version-history":[{"count":1,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2635\/revisions"}],"predecessor-version":[{"id":2636,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/posts\/2635\/revisions\/2636"}],"wp:attachment":[{"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/media?parent=2635"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/categories?post=2635"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/howtogeek.blog\/lv\/wp-json\/wp\/v2\/tags?post=2635"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}