Langue
Affichage
Grille d'images
Autocomplétion du prompt
Filtrage du contenu
icône de réclamation quotidienne : verre vide
Réclamation quotidienne
Aujourd'hui
D
+3
L
+3
M
+4
M
+4
J
+5
V
+5
S
+6
Réclame chaque jour pour faire grandir ta série.
Art IA
11

Image to Image Test

Wanted to test the Image to Image on a sketch and see if this would complete it like other language models. This is the result at 50% with the prompt shown. The sketch I did at work on a yellow note pad and you can see it coming through with the transparency shown on the actual character and the lines from the notepad making up the "rooftop." Looking at the thumbnail you can almost tell it was a yellow pad lol. Pose I can give 9/10 for this test. The anatomy and proportions of the sketch came across almost perfectly. The sketch was pretty barebones on details other than construction lines, not even a face. Coming up with the rest based on the word "ninja" is quite promising. If only I was good at more NSFW poses, this could really elevate what I could achieve here 🤣😭 Have any of you tried Image to Image for just a rough pose of what you wanted? What do you guys think of this result? Wish I could show you guys the actual sketch @terror2 @warmicestudios @gildenstein @peterco @bad77 @zuzul @cef_ultra

Données de génération

Prompt
complete this sketch of a ninja, on rooftop, night, holding swords, manga art, toriyama akira (style), colored pencil (artwork),
Prompt négatif
text, signature, tan lines,
Seed
1092174577
Steps
28
CFG Scale
5
Sampler
Euler a

Plus debarry915

Work Adventures - Rōze: DrummerWork Adventures - Victoria: AttorneyWork Adventures - Hime: Parking Lot AttendantWork Adventures - Penny: MaidWork Adventures - Sarah: Pizza Delivery GirlWork Adventures - Hana: Nurse

Commentaires (11)

G
Yeah tried it out already 😁 my Rockband shot was created this way 😉
P
This one looks cool. I experimented with ref. images too, in Amanatsu only, though. At 50% it is necessary to keep correct ratio, otherwise it could be stretched or deformed. Oh, and there might be distorted fingers 👍
W
Great job Barry! Saw the original image you shared and I think you are on the right track!! — Cool detail that the actual lines on the pad became the rooftop tiles!!!!! 🤣🤙🏻
G
It actually took me many attempts to get it right... try and error 🤣
T
I haven't tried tis yet. Mainly because I can't draw. I don't think the AI can work with stick figures. 😆
B
😈: very promising!! It’s gotten the pose down really well. The Yellow paper effect actually makes it look more retro, I like it. I haven’t used image ref on sketches yet. I’ve just used them on Annie’s photos. It gets her poses mostly right on 100%. But I think it definitely struggles more with pictures than sketches.

Posts similaires