Newsmeldung von slashCAM:
Das neue Feature "Style Tuner" bringt Midjourney einen großen Schritt näher zur konsistenten Erzeugung von KI-Bildern mit einem durchgehenden Stil. Die hierfür notwendige...
Hier geht es zur Newsmeldung auf den slashCAM Magazin-Seiten:
Durchgehende Stile in Midjourney definieren und mixen - Style Tuner
Antwort von Alex:
Midjourney's Consistent Character feature is MASSIVE.
For any/all story tellers.
Storyboarding became easier (overnight)
Characters stay consistent in...
+ Multiple Settings
+ Multiple Poses
+ Multiple Actions
So if you have a story to tell…
You can now do it with MJ.
That being said…
Prompting correctly is essential.
Prompt Tips:
+ Be extremely specific
+ Focus on very small details
+ Like eye direction/body part position/etc
+ Reinforce clothing, hair, etc
+ Build a story with your words
+
***Disclaimer:
This is a very simple storyboard.
You don’t just push one button...
And get a finished. (Needs to be said)
Here’s how to do it:
01. Create a base character in MJ
02. Copy image url
03. Create new prompt
04. Add parameter
05. Paste image URL
06. Generate images.
07. Iterate prompts.
Base Prompt structure:
Photo type: street photography, , subject:, ,, wearing: subject focus: sharp focus, mid-ground, environment: , , texture: , image type: 35mm --ar 2:1
Prompt Example 1:
Photo type: street photography lifestyle, closeup side profile, subject: 30 year old man, sitting on front porch steps, sipping coffee from a small white mug, mug lifted to his mouth, mouth on mug, subtle steam from the mug, tan skin, dirty blonde hair, wearing: lime green beanie, oversized acid wash grey hoodie, black jeans, subject focus: sharp focus, mid ground, environment: mountain environment, bright morning light creating rim effect, texture: minimalist clean architecture, image type: 35mm --ar 2:1 --cref {image link}
Prompt Example 2:
Photo type: street photography lifestyle, closeup diagonal shot, subject: 30 year old man, putting on a lime green beanie, hands up, pulling sides of beanie onto his head, tan skin, wet dirty blonde hair, beard stubble, subject focus: sharp focus, mid ground, environment: modern bedroom, diffused lighting, texture: minimalist clean architecture, image type: 35mm --ar 2:1 --cref {image link} --cw 90
Thoughts:
+ This is not perfect
+ Tattoos will never be consistent
+ This story is super basic
+ But you can see how this could apply broadly.
https://www.linkedin.com/posts/rory-fly ... er_desktop
1710367053748.jpeg
Antwort von Frank Glencairn:
Ich hab die erste Prompt mal spaßeshlaber in verschiedene AIs geschmissen:
image_2024-03-15_122856580.png
Freepik
image_2024-03-15_123002079.png
Leonardo
image_2024-03-15_123108089.png
Civitai
image_2024-03-15_123138734.png
Tensorart
image_2024-03-15_123247146.png
Magespace
image_2024-03-15_123312667.png
NightCafe
image_2024-03-15_123541970.png
openart
image_2024-03-15_123616851.png
Seart
Bottom Line: Midjourney ist immer noch kaum zu schlagen, einzig Civitai kommt auch nur annähernd dahin.
Antwort von Frank Glencairn:
Und just for fun Vlad der Pfähler, Clark Kent, Hitchcock und Luke Skywalker :D
image_2024-03-15_124934188.png
image_2024-03-15_125148418.png
image_2024-03-15_125719888.png
image_2024-03-15_125651472.png
