A: Thanks for raising this — we understand how frustrating that experience can be, especially when you're aiming for something specific like a bald look.

Here’s how to get more consistent results, especially for appearance-based features:

✅ Step 1: Use consistent training images
The model is specifically optimized to map facial structure and identity with the highest accuracy. So, for features like baldness, it’s important that all uploaded images reflect the same visual — fully bald in this case. Even a slight inconsistency (e.g., stubble, lighting reflections, shadows) can lead to the model generating hair where you don’t want it.

✅ Step 2: Guide the look with the prompt
The rest of the appearance — including hairstyle, body type, outfit, setting — is controlled through presets or custom prompts. Since presets might generalize, we recommend overriding with a descriptive text prompt like:
“Confident bald man with a clean-shaven head, standing under natural light, wearing a fitted black t-shirt.”

This lets the model stay true to your facial identity while adopting the rest of the visual style you specify.

We're constantly working to reduce such inconsistencies and improve trait control. Let us know how this approach works for you — and happy to troubleshoot further if needed!