Stability AI tau tshaj tawm cov qauv kev kawm tshuab hu ua Stable Video Diffusion uas tuaj yeem tsim cov yeeb yaj kiab luv luv los ntawm cov duab. Tus qauv nthuav dav lub peev xwm ntawm Stable Diffusion project, yav tas los txwv rau kev sib txuas ntawm cov duab zoo li qub. Cov cai rau kev cob qhia neural network thiab cov cuab yeej tsim duab yog sau hauv Python siv PyTorch lub moj khaum thiab luam tawm raws li MIT daim ntawv tso cai. Twb tau kawm ua qauv tau qhib raws li kev tso cai Creative ML OpenRAIL-M daim ntawv tso cai, tso cai rau kev lag luam siv.
Muaj ob txoj kev xaiv muaj rau rub tawm: SVD (Stable Video Diffusion) rau tsim 14 thav duab nrog kev daws teeb meem ntawm 576x1024 raws li cov duab zoo li qub thiab SVD-XT rau tsim 25 thav ntawv. Nws muaj peev xwm tsim cov yeeb yaj kiab yam tsis muaj kev txav los yog nrog lub koob yees duab qeeb heev, kav ntev tsis tshaj 4 vib nas this. Kev tswj tus qauv ncaj qha raws li cov lus piav qhia ntawm cov lus piav qhia tseem tsis tau txais kev txhawb nqa, tab sis koj tuaj yeem ua ntej npaj cov duab qub uas siv tus qauv qub Stable Diffusion 2.1 thiab tom qab ntawd hloov mus rau video siv tus qauv SVD.
Video zoo tseem tsis tau muab qhov zoo tagnrho photorealism thiab lav qhov tseeb rendering ntawm lub ntsej muag thiab tib neeg. Nyob rau hauv cov nqe lus ntawm kev ua tau zoo, lub tswv yim qhib qauv yog ua ntej ntawm proprietary analogues los ntawm Runway thiab Pika Labs. Cov qauv tuaj yeem hloov kho tau yooj yim los daws cov teeb meem ntau yam, piv txwv li, nws tuaj yeem siv los ua cov duab peb sab.
Tsis tas li ntawd, peb tuaj yeem sau tseg qhov kev tshaj tawm ntawm Cov Khoom Siv Video-LLaVA tshuab kev kawm, uas tso cai rau koj los tsim ib qho kev sib koom ua ke pom kev sawv cev ntawm ib qho khoom, tsim los ntawm kev siv cov duab ib txhij thiab cov ntaub ntawv kaw tseg ntawm cov khoom thaum kawm. Lub kaw lus siv tau, piv txwv li, kom paub txog qhov muaj tib yam khoom hauv cov duab thiab yeeb yaj kiab. Cov cai tau sau hauv Python thiab muab faib raws li Apache 2.0 daim ntawv tso cai.
Tau qhov twg los: opennet.ru
