118 attributes, one face. A technical look at how we generate the same face every time using DNA-like attributes.
The 118 Attributes
Instead of a free-form text prompt, we use a structured DNA object. Each attribute—ethnicity, age range, gender presentation, facial structure, skin tone, hair color, eye color, and many more—is set once and reused for every generation.
Same DNA input → same face output. Every time.
- → Ethnicity, age, gender presentation
- → Skin tone, hair color, eye color
- → Facial structure, bone structure, features
- → Locked across platforms and regenerations
Why Structure Beats Prompts
Free-text prompts are ambiguous. "East Asian woman, 25–30" can yield a different face on each run. Our DNA schema removes that ambiguity: the model receives the same structured vector every time, so the output stays consistent.
What You Get
When you create an identity with the Agent API, you get a stable image URL. Use it as your avatar everywhere. Regenerate with the same DNA as often as you like—the face stays the same.