Modèlu diffusion wideo
Modèlu diffusion wideo yi dañuy defar ay nataal yuy toxu ci di soppi bruit yu bari yi ci kadre yu lëkkaloo, di yokk xalaatu diffusion bi ci nataal yi dem ci jamono.
Résumé
They are the engine behind today's most realistic AI video.
Plongeur bu xóot
Modèle diffusion yi dañu jàng ni ñuy delloosi bruit bi: ci diiru tàggat yaram, done yu sell yi dañuy yokk bruit ndànk-ndànk, ba noppi reso bi dafay jàng xam luy waaja am ak dindi bruit boobu jéego par jéego. Diffusion wideo dafay jëfandikoo lii ci toppalante kadre yi, ak yokk bu am solo ci modeling temporel suko defee mouvement bi nekk bu nooy te mbir yi di wéy di nekk benn ci diir bi. Ngir mëna yombal xayma yi, sistem yu bari ay modelu diffusion yu nëbbu lañu, ñuy dox ci barab bu nëbbu buñ komprime, du ci pixel yu ñor. Architecture yi dañu tàmbalee ci 3D U-Nets ak bàyyi xel ci wàllu barab ak jamono ba ci transformatëri diffusion (DiTs) yuy jàppee wideo ni ay token barab-jamono. Famille bi dafay doxal Sora, Diffusion wideo bu dëgër, piste Gen-3, Google Veo, ak Pika, ba noppi jàppale bind-ci-wideo, nataal-ci-wideo, ak soppali wideo.
Gis-gis xarala
Li gëna am solo mooy yokk ay couche temporel, lu melni attention temporelle wala convolutions 3D, suko defee kadre yi di denoisé ñoom ñaar, duñu denoise seen bopp, loolu mooy tere flicker ak mouvement incoherent. Generation dafay jëfandikoo tegtal bu amul benn classifier ngir topp liñu bind ci kàddu yi, ba noppi encoder/decoder VAE bi jàng dafay toxu diggante pixel yi ak barab bu nëbbu bi. Sampling jéego yu bari yu denoising yeexna, moo tax distillation ak solver yu gëna gaaw lañuy jëfandikoo ngir wàññi limu jéego yiñ soxla.
njeextalu pexe
Gaawaay ak yaatuwaay
Visual IA mën na otomatise saytu, gis ak etiketu liggéey ci eskaal.
Tabax tànneef
Ekipu kreatif yi mën nañu defar konsept yu gëna gaaw te duñu def lu bari ci loxo.
Ekip ak def liggéey
Liggéeyukaay yi mën nañu jëfandikoo siñaal nataal wala wideo yu jafewoon lool ci liggéey.
Ëlëgu modelu diffusion wideo
Gëstu bi mingi daw ngir gëna gudd, gëna am dayo, defar ci jamono dëgg, ànd ak audio buñ boole ak realisme physique bu gëna baax. Transformatëri diffusion yiy eskaleer bu baax ak done ak xayma ñooy gëna am solo, ba noppi model distiye yu néew jéego ñoo gëna gaaw ci defar. Xaarandil ñu gëna mëna saytu kamera yi, personage yi, ak soppali yi, boole ci jegewaale yu boole diffusion ak yeneen pexe yu defar. Lu kalite bi di gëna yokk, watermarking bu dëgër ak ëmbiit li-provenance dafay nekk lu am solo ngir moytu jëfandikoo ludul yoon.
Doxal ci àdduna dëgg
Jumtukaayi bind-ci-wideo yu am doole yu melni diffusion wideo bu dëgër, Piste Gen-3, ak Pika ngir soskat yi
Animaasioŋ nataal-wideo buy dundal benn nataal ak yëngu-yëngu bu dëggu
Soppi wideo ci ndimbalu IA, peinture, ak toxal stil ci biir liggéey bi ginaaw defar
Defar ay wideo tàggat yaram ak ay simulaasioŋ ngir robotik ak gëstu ci oto yuy demal seen bopp
Risk yi ak balustrade yi
Yelleefi nataal ak nangu mën na nekk risku yoon sudee fi ñu bawoo leerul.
Performance model bi mën na wuute ci leeraay bi, demographie bi ak environmaa bi.
Njuumteg positive yi mën nañu dem te kenn duko seetlu fileek xool wuñu buntu wóolu sa bopp.
Roadmap ngir samp gi
Mandargal kritërium nangug njub, woowaat ak njëgu njuumte.
Saytu ak done yu méngoo ak anam yi ñuy liggéeyee dëgg.
Yokk jàngat nit ngir xam fu wóorul dara wala am njeexital yu rëy.
Toppal model drift bi nga baaxal ko ginaaw bi kamera bi wala done yi soppeekoo.
Weyal di banneexu
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 Video Diffusion Models 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
Gis bi ci topp
Modèlu diffusion GLIDE
Laaj yi ñuy faral di laaj
What is Video Diffusion Models?
Modèlu diffusion wideo yi dañuy defar ay nataal yuy toxu ci di soppi bruit yu bari yi ci kadre yu lëkkaloo, di yokk xalaatu diffusion bi ci nataal yi dem ci jamono. Ñoom ñooy motër bi ci ginaaw wideo IA yi gëna dëggu tay.
Lan mooy xalaat bu am solo bi ci ginaaw xeetu diffusion?
Modèle diffusion dañu leen di tàggat ngir dindi bruit bi ñuy yokk ndànk ci done yi, ba noppi ñu defar ci denoising ci bruit bu sell bi.
Lan la xeetu diffusion wideo yokk ci xeetu diffusion nataal?
Ginaaw defar kadre bu nekk, diffusion wideo dafa wara boole kadre yi ci diir bi, te loolu dafay laaj ay couche temporale ngir mëna wéy di doxal.
Lan moo waral modelu diffusion wideo yu bari di doxal ci barab bu ' nëbbu'?
Wideo Raw dafa rëy; encoder ko ci barab bu gëna ndaw bu nëbbu jaaraleko ci VAE dafay tax denoising gëna am njariñ.
Lan mooy wareefu bàyyi xel ci jamono wala convolution 3D ci xeetu mbir yooyu?
Couche temporale yi dañuy boole xibaar yi ci biir kadre yi, di tere flicker ak defar mouvement bu méngoo, du kadre yu moom seen bopp, jittery.
Ban pexe mooy jàppale modelu diffusion wideo mu topp ab bind bu dëgër?
Njàngale bu amul klassifikatër dafay gëna dooleel denoising bi ci wàllu kondisioneer, gëna baaxal topp gi.