SDXL da Cascaded Diffusion
SDXL shine babban ƙudurin rubutu-zuwa hoto samfurin Stability AI wanda ke haɗa babban janareta mai ƙarfi tare da mai tacewa, yayin da ɗimbin yaɗuwar sarƙoƙi yana ɗaure samfura da yawa don gina hotuna daga ƙarami zuwa babban ƙuduri.
Dubawa
Together they explain how modern open-source image generators hit photorealistic quality.
Zurfafa nutsewa
SDXL (Stable Diffusion XL) shine tsarin watsawa na kusan biliyan 3.5 wanda ke samar da hotuna 1024 × 1024 a asali, babban tsalle akan 512x512 na asali Stable Diffusion. Yana amfani da maɓallan rubutu guda biyu (OpenCLIP ViT-bigG da CLIP ViT-L) don ƙarin fahimtar saurin fahimta, da girma da yanayin amfanin gona don haka ƙirar ta san ƙudurin manufa da ƙira. SDXL yana jigilar bututun mai mataki biyu: ƙirar tushe tana haifar da hoton ɓoye, sannan samfurin mai tacewa na zaɓi yana ƙara dalla-dalla dalla-dalla a cikin matakai na ƙarshe. Yaɗuwar da ba a taɓa gani ba shine mafi girman ra'ayin da ke bayan wannan: maimakon ƙira ɗaya yin komai, kuna sarƙar ƙaramin ƙirar ƙirƙira ƙaramin hoto mai ƙarancin ƙuduri tare da ƙirar yaduwa mai ƙarfi waɗanda ke haɓaka shi, kowannensu ya horar da matakinsa. Hotunan Google ya shahara da tsarin cascade.
Fahimtar Fasaha
Dukansu biyu suna aiki a cikin tsarin ɓarna: farawa daga hayaniyar bazuwar da tsinkaya akai-akai da cire shi, ta hanyar rubutu. SDXL yana aiki a cikin madaidaicin sarari ta hanyar VAE, don haka ƙididdigewa yana da arha fiye da yin aiki akan ƙananan pixels. Mai tacewa wani keɓantaccen samfurin ƙwararru ne wanda ke ɗaukar matakai na ƙarshe, ƙananan amo. A cikin kasike na gaskiya, ƙirar tushe tana fitar da ƙaramin hoto, sannan samfuran yaɗuwar yanayin babban ƙuduri sun haɓaka shi, kowanne yana yin sharadi akan ƙaramin ƙuduri, galibi yana amfani da ƙarar yanayin ƙarar amo don tsayawa tsayin daka.
Dabarun Tasiri
Gudu da sikelin
Kayayyakin AI na iya sarrafa aiki da bincike, ganowa, da ayyuka masu alama a sikelin.
Gina zaɓuɓɓuka
Ƙungiyoyin ƙirƙira za su iya samar da ra'ayoyi cikin sauri tare da ƙarancin bita da hannu.
Ƙungiya da aikin aiki
Ayyuka na iya amfani da siginar hoto da bidiyo waɗanda a baya suke da wahalar aiwatarwa.
Makomar SDXL da Yaɗuwar Cascade
Halin yana zuwa ƴan ƙasa kaɗan, matakai masu sauri da haɗin kan gine-gine. Hannun distillation kamar SDXL Turbo da Samfuran Daidaitawa sun riga sun yanke tsara zuwa matakai ɗaya zuwa huɗu. Masu watsa shirye-shiryen watsawa (kamar yadda yake a cikin Stable Diffusion 3 da FLUX) galibi suna maye gurbin U-Net kashin baya, kuma tsararru mai ƙarfi na ƙarshen zuwa-ƙarshen yana rage dogaro ga fayyace cascades. Yi tsammanin haɗaɗɗiyar gyare-gyare, ingantacciyar ma'anar rubutu, da haɗin hoton kan na'ura na ainihi yayin da inganci ke ci gaba da inganta.
Aiwatar da Gaskiyar Duniya
Samar da tallace-tallace na 1024x1024 da zane-zane kai tsaye daga faɗakarwar rubutu ba tare da keɓancewar sama ba.
Yin amfani da bututun tushe-da-mai tacewa SDXL don ƙara dalla-dalla ga fuska da laushi a cikin izgili na samfur.
Gudun SDXL Turbo don samfotin hoto na kusa-kusa a cikin kayan aikin ƙira na mu'amala
Gina babban kascade na al'ada don juyar da ƙananan zane-zane zuwa manyan hotuna masu ƙarfi.
Hatsari & Tsare-tsare
Haƙƙoƙin hoto da yarda na iya zama haxarin doka idan ba a fayyace ba.
Ayyukan samfuri na iya bambanta a ko'ina cikin haske, ƙididdiga, da mahalli.
Ƙarya tabbataccen ƙila ba za a iya lura da shi ba sai dai idan an kula da ƙofofin amincewa.
Taswirar Hanya
Ƙayyade ma'auni na karɓa don daidaito, tunowa, da farashi na kuskure.
Gwada tare da bayanan da suka dace da ainihin yanayin samarwa.
Ƙara bita na ɗan adam don ƙarancin amincewa ko tsinkaya mai tasiri.
Bi diddigin ƙirar ƙira kuma sake ingantawa bayan canje-canjen kamara ko saitin bayanai.
Ci gaba da Bincike
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 SDXL and Cascaded Diffusion 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
Jagora na gaba
Tuning Multi-Concept Tuning
Tambayoyin da ake yawan yi
What is SDXL and Cascaded Diffusion?
SDXL shine babban ƙudurin rubutu-zuwa hoto samfurin Stability AI wanda ke haɗa babban janareta mai ƙarfi tare da mai tacewa, yayin da ɗimbin yaɗuwar sarƙoƙi yana ɗaure samfura da yawa don gina hotuna daga ƙarami zuwa babban ƙuduri. Tare sun bayyana yadda na'urorin buɗaɗɗen hoto na zamani suka buga ingancin hoto.
Wane ƙuduri na asali SDXL ke samar da hotuna a, idan aka kwatanta da asalin Stable Diffusion?
SDXL a asali yana samar da hotuna 1024x1024, yanki mai girma sau huɗu fiye da na asali Stable Diffusion's 512x512.
Menene aikin samfurin matatar mai SDXL?
Mai tacewa wani keɓantaccen samfurin ƙwararru ne wanda aka yi amfani da shi a ƙarshen bututun don haɓaka daki-daki bayan ƙirar tushe ta ƙirƙiri hoton ɓoye.
Rubutun rubutu nawa SDXL ke amfani da su?
SDXL yana amfani da maƙallan rubutu guda biyu, OpenCLIP ViT-bigG da CLIP ViT-L, waɗanda aka haɗa don ƙarin fahimtar saurin fahimta.
Menene ainihin ra'ayin yaɗuwar cascade?
Yadawa sarƙoƙi yana ɗaure ƙaramin janareta na tushe tare da nau'ikan yaɗawa mafi girman ƙuduri ɗaya ko fiye, kowanne yana da sharadi akan mafi ƙarancin ƙudurin baya.
Me yasa SDXL ke ƙi a cikin ɓoyewar sarari maimakon kai tsaye akan pixels?
A VAE yana matsa hotuna zuwa cikin ƙaramin sarari mai ɓoye, don haka tsarin watsawa ya fi arha fiye da aiki akan ƙananan pixels masu cikakken ƙuduri.