[Mrs. Yor Cosplay] Genshin Impact AI Figure Generation Tutorial Sharing

Using AI to transform cosplay photos into physical figure models offers a cross-dimensional fusion that lets you experience the sheer fun of being a creator. Here, I'm sharing a complete creative workflow and physical display.

First comes the prep work: shooting a crisp, well-lit cosplay photo. Since this experiment combines elements of Mrs. Yor cosplay with a Genshin Impact aesthetic, I created a fused secondary design for the props and outfit styling. After completing the base photo shoot, it's time to use the Lmarena.ai tool. Open the main page, click the plus icon to upload the chosen photo, enter the Image module, and paste a specified prompt code. The code specifies using the nano-banana model to craft a 1:7 scale commercial figure model and sets up the specific desktop display environment for the final product.

The visual composition of the AI-generated output is very intriguing. The figure model is placed on a computer desk, set upon a completely textless, transparent circular acrylic base. On the computer screen behind the figure, the Character modeling wireframe process in ZBrush is accurately rendered—this combination of digital modeling and physical display significantly elevates the visual layering. Another highlight is the Bandai-style toy packaging box placed beside it, printed with key art identical to the original image. The final AI-generated figure model pays high attention to facial details, striving for an extraordinarily photorealistic recreation of the original cosplay photo. Concentrating complex Character modeling processes, packaging design, and physical visual effects into a single desk setup scene imbues the composition with craftsmanship and a collectible atmosphere.

Tracing back to the actual offline cosplay creation, the process was far from easy. Selecting materials for this blue-and-white armored outfit took considerable effort, utilizing a patchwork of leather and stiff fabric, with the back cape relying on stiff interfacing to hold its voluminous shape shown in photos. Testing patience most was the matching large weapon prop: hand-sculpted and sanded from EVA foam, coated in a gradient ice-blue paint finish, and accented with gleaming metallic gold hardware. The footwear features tactical knee-high boots; to elongate the visual standing proportion, the sole thickness was specially tweaked, which, along with thigh harness straps, rendered leg lines sleek and tall. For the wig, light-blue heat-resistant fiber was trimmed and styled into crisp, layered L-shaped bangs. The makeup adopted a cool-toned palette with swept-up red eyeshadow at the outer corners, capturing the character's signature blend of cool elegance and soft grace.

Comparing the 6 original images side-by-side, the online AI-generated model and offline physical shoot form a wonderful synergy. The first few images feature mesh models on screen, a desktop ornament holding a red umbrella, a standing figure in a tight black dress, a seated character in a pink robe, and a red-haired character figure model in formal attire holding a cup. These diverse standing and sitting posture samples offered rich style references for AI training. Meanwhile, my raw photo was captured against white drapes under mixed warm-and-cool lighting—a blurred background coupled with a closed-eye close-up and a relaxed pose holding a broadsword directly served as the data source for the final digital model.

Digitizing 2D flat cosplay photos honestly gives a genuine sense of breaking through the fourth wall, as if the entire creative journey gains a 3D physical extension. If time permits, I hope to continue exploring converting more characters into figure models on this platform, unlocking possibilities of merging different costume textures, action poses, and AI-generated workflows. Whether displayed on a desk setup or shared on social media, the final output possesses a deeply enduring texture.