- Using low-quality images for training.
- Insufficient images focused on your face or upper body, or an overabundance of images that do not prominently feature your face.
Include suggested tags in your prompt for more accurate results. For instance, if your images feature tattoos or lipstick, we recommend adding those keywords (e.g., "tattoos, lipstick") to the prompt. Without them, the AI might overlook these details, resulting in images that differ slightly from your appearance.
We suggest experimenting with the "regenerate face with my trained model" option on the generation page, toggling it on and off to determine what yields the best results for you.
Keep in mind that our model trainer, while designed to be user-friendly, is not perfect. Since AI involves probabilistic processes, we cannot guarantee flawless results every time. However, we strive to make it as straightforward as possible for anyone to train, use, and even monetize their models without requiring technical expertise.