Yes! People who know me saw this coming already even before I started this blog. Being a massive Disney Frozen fan, my ultimate goal in this whole A.I. exploration adventure, is to create my own Anna and Elsa images… obviously…
So, today I am going to try to have some of the starter models recreate this image for me:
For this, I will feed all models the same prompt:
- General Composition & Aesthetic
Shot Type: Close-up duo portrait / Upper torso shot. The subjects are posed closely together, back to back, slightly overlapping, facing forward with gentle, warm expressions.
Background: A solid, deep royal blue backdrop (like a painted canvas or textured velvet curtain) with faint ice crystal patterns. Crucially, the background must remain simple and non-distracting.
Mood/Tone: Ethereal, whimsical, romantic, and highly polished; typical of a major Disney movie poster advertisement.
Lighting: Professional studio lighting setup (softbox diffusion). The lighting is high-key, meaning bright overall with very soft, gentle shadows. There must be distinct sparkle and glow highlights catching the hair and dresses.- Subject Details: Elsa (Left Character)
Hair: Voluminous platinum blonde hair. Styled in a graceful, elevated half-up/half-down manner, allowing for maximum volume around the crown and shoulders. The texture should be flawlessly styled but appear natural.
Eyes & Makeup: Large, luminous eyes colored bright ice blue (or lavender/periwinkle). Immaculate makeup: subtle pink shimmer shadow across the lids, defined eyeliner, and perfectly sculpted eyebrows.
Attire: A sparkling gown in a light aqua or pale icy blue color. The fabric must look heavily sequined, beaded, or crystal-encrusted on the bodice, giving it maximum reflective shine under the studio lights.
Expression: Serene, poised, graceful, with a soft and enigmatic smile.- Subject Details: Anna (Right Character)
Hair: Rich auburn reddish-brown hair. Styled in two thick braids that cascade over her shoulders. The color must have natural depth and variation (not solid orange).
Eyes & Makeup: Bright, warm sapphire blue eyes. Her expression is open, joyful, and enthusiastic, looking slightly past the camera towards her sister Elsa with a wide, genuine smile.
Attire: A black summer dress covered in rosemaling patterns, covered at the shoulders by a velvet-purplish cloak with similar patterns around the rims.
I will create 6 images on each model and select the one I think resembles the original the closest. All 6 images will also be available to those interested…
First run, FLUX.1 dev using a 640×640 canvas, 30 steps, 4 guidance, random seed…
FLUX.2 Klein is a small and very fast model that does a nice job, although it doesn’t seem to be too good at doing animation style images…
Stable Diffusion 1.5 / Dreamshaper 8, 30 steps, 7 CFG Scale… Clearly not suitable for this experiment. It looks nothing like I intended…
Anima… also not really what I was hoping it would be… In fact, it is almost creepy. It is supposed to be a model for cartoon-style images, but it looks like it trying its best (and failing) to make it look realistic. Maybe there is something wrong with my prompt. There’s also not much variation…
Z-Image… very little variation. But it does follow the prompt quite nicely.
Qwen is very large model of 58 GB and I am very confused for it to actually run at all on my system, but that is going to be a topic for a (near) future post. As for my Anna and Elsa prompt, I think it is one of the better ones!
Leave a Reply