//=time() ?>
stablediffusion2.0ではsteps低めでもいけるらしいので実験。DrawThingsでの最小設定10でも「絵」は出来る(低いほど速い、デフォルトは30)。
Guidance Scale(プロンプトに忠実になる)のデフォルトは7.5。いつもMAXの25でやってたけどsteps10にすると残念なことに。
各画像の設定はALT参照。
@Biz_Horse Drawn by StableDiffusion2, Style text added.
Full Prompt: https://t.co/9Nln1Ij2q2
AD: Search for Hat on Amazon https://t.co/ojgQ3Cd81q
#StableDiffusion2 & textual inversion. 2.0 really does seem great for customization
@sonicx246 Drawn by StableDiffusion2
Watermark Free: https://t.co/9tISqGRYDf
AD: Search for Hairstyle on Amazon https://t.co/eNZ9vWTVPA
@RaionLion_ Drawn by StableDiffusion2, Style text added.
Full Prompt: https://t.co/JKkqrSmcHM
AD: Search for Vehicle on Amazon https://t.co/K1jiRUCpUa
@Devilled_MP3s Drawn by StableDiffusion2
Watermark Free: https://t.co/bNrDDCv5Po
AD: Search for Hat on Amazon https://t.co/ojgQ3Cd81q
@UareSuchA Drawn by StableDiffusion2
Reply Commands: https://t.co/91ADKkk14L
AD: Search for Fashion Accessory on Amazon https://t.co/Mrm6lr6Ts4
@2kplaye19874842 Drawn by StableDiffusion2, Style text added.
Full Prompt: https://t.co/4UZIk860z5
AD: Search for Audio Equipment on Amazon https://t.co/Fa1z2iv8Z0
@Juanj348 Drawn by StableDiffusion2, Style text added.
Reply Commands: https://t.co/AlykVbQfRU
AD: Sponsor This Account. See https://t.co/QtotRaK0Wj
I finally finished a v2 training and it copies the style VERY well. But almost too well? With my 1.5 training, I was able to morph old Disco styles into other styles while maintaining their essence.
here its DD but a little worse
Keeping at it.
#StableDiffusion2 #DeepLearning
@Targamos777 Drawn by StableDiffusion2, Style text added.
Watermark Free: https://t.co/xU4Wg0jXW7
AD: Search for Hairstyle on Amazon https://t.co/eNZ9vWTVPA