Model Swap Input Requirements
For our Model Swap feature, you provide an image of an existing model, and the AI replaces their face and body identity while keeping the garment and pose. To achieve the highest quality and most consistent results, the image you provide must meet the following specifications:
Image quality and composition
Section titled “Image quality and composition”- Single subject only: The input image must contain only one model.
- Uncropped model: The model in the image must not be cropped (their full body should be visible within the frame).
- Minimum resolution: Images must have a minimum resolution of 1.5K (e.g., 1500 pixels on the shortest side). Higher resolution always leads to better results.
Identity matching guidelines
Section titled “Identity matching guidelines”To prevent visible artifacts and maintain realism, the characteristics of your input model should align closely with the characteristics of the target identity you select from our system:
| Input attribute | Required match for target identity |
|---|---|
| Morphology | The input model’s body shape (e.g., height, build) must be similar to the target model’s morphology. |
| Gender | The gender of the input model and the target identity must match. |
| Age | The input model should be in a similar age bracket (e.g., adult). |
| Skin tone | The input model’s skin tone should be reasonably similar to the target identity’s skin tone for consistent results. |
| Hairstyle | Avoid input images with highly complex, intricate, or unusual hairstyles that may conflict with the target identity’s default hair, as this can lead to unnatural blending. |