Designing a Game Wardrobe: Essential Tips for Creators

Master the art of outfitting avatars with style, consistency, and precision, ensuring their attire is both appealing and accurate in every iteration

The demonstration utilizes stylish images sourced from Pexels, an open-source photo platform. The primary objective is to achieve consistent clothing representation across various outfits and characters, ensuring uniformity in design and style.

The walkthrough will concentrate on styling one character, providing a clear example of the process. However, this is just a glimpse of a broader range of fashion possibilities. Detailed, step-by-step instructions will be provided for replicating these styles, focusing on a repeatable workflow that can be applied to a diverse lineup of characters and outfits. This approach ensures users can achieve precise and consistent results in their avatar customization.

Customization is key for achieving a specific look, as evident in the side-by-side comparison. This demonstrates the limitations of basic prompts, highlighting the significant impact of detailed customization in realizing a desired aesthetic.

 

Ensure you have a custom character prepared for this fashion customization process. If you haven't created one yet, click here for an easy-to-follow guide to develop your own.

Step 1

Select your reference images carefully. Experiment with using 5-20 images per outfit, with 15 being optimal for capturing intricate outfit details. Crop out the heads from the images to maintain focus on the outfits (apologies to the models!). For optimal results in SDXL training, resize the pictures to double the recommended size of 1024x1024. This allows for zooming and cropping later without compromising the details and quality of the fashion elements.

Step 2

Make sure to use a unique trigger token for the piece of clothing when you caption your dataset.  For a quick refresher on this, refer to this article about manually captioning your dataset. When dialing in the details for your model, consider these recommended settings and captions:

• Training Steps: Multiply your image count by 350.
• "Subject" Setting to Maintain focus on the Object
• Unet Learning Rate: Set to 5e-5
• Text Encoder Ratio: Opt for 0.35
• Text Encoder LR: Use 1e-5

Step 3

While the outfit model processes, prepare the base pose using the your unique character model.. This highlights adapting outfits to diverse body types. In our example, we use the Gwendolyn character model, and her pose shows most of her frame, focusing less on shoes (a full A-Frame pose would be used for shoe emphasis). View the specific prompts for this setup in the provided screenshot.

If your models are trained then you are ready to make a composition with your character and outfit models together. The animated image below shows a quick look at the process but check out article here if you need a refresh on Composing Multiple LoRAs. 

Step 4

Next, it's time to update the base image, the character's pose we generated in step 3. Locate your base image in the gallery, and select 'Refine'.

This leads to the generation screen with your image and the initial prompt. Add to the beginning of your prompt “wearing [unique token and outfit description]” for better precision.

Use the sketch feature to color block on your character. This is essential for providing accurate color cues to the model. Use the eyedropper tool to match colors directly from your reference photos, ensuring attention to detail. If you need help accessing the sketch tool click here.

Be aware that faces may slightly alter during the styling process. This is a normal occurrence, but there's a straightforward solution which we will show you in the next step.

Cick on your favorite outfit variant to open the image view page, then click 'Send to Canvas'.

Step 5

In the video below, we'll show you how we reintegrate the original base image of the pose we made in Step 3. This involves resizing and swapping heads to ensure a perfect match, along with using the masking tool for fine-tuning. The canvas may require some time to process, so consider using this interval to start styling your next look. 

This stage allows for creativity – feel free to adjust hairstyles, crop, repaint, or make any modifications that enhance your fashion design. Each adjustment contributes to the uniqueness and appeal of your character's outfit. 

 

 

Gwendolyn's outfit exudes a springtime essence, featuring a soft pink dress adorned with subtle floral prints. Its flowing design and sweet, delicate patterns encapsulate a relaxed charm, perfectly blending casual comfort with a touch of elegance. This look is ideal for sunny, laid-back days, capturing the essence of effortless style.

 

Gwendolyn's latest outfit features a twist on the classic trench coat. In a traditional beige, the coat combines tailored lines with softer textures underneath, offering a mix of sophistication and a touch of whimsy. Ideal as a transitional piece for shifting seasons, it's given an unexpected spring flourish with the addition of a flower crown.

 

Gwendolyn makes an entrance in a bold coral one-piece swimsuit, embodying the essence of summer. Its simple yet striking design captures the 'sunshine and chill' mood, making it ideal for beach days or poolside lounging. Just add water, and she's ready to make a stylish splash!

 

In a setting where the horizon is vast, Gwendolyn showcases a desert-chic, festival-ready look. Featuring high-waisted shorts and a breezy black crop top, the outfit exudes a cool, relaxed vibe. It's the ideal attire for long summer days that transition into starlit nights.