I tested the product on an a selfie and my deduction is that for a free product it is doing pretty good but have you compared it with Gemini's free image generation models. I think when compared with Gemini as a standard, it is coming a little off-target, although its just my opinion and the reason being is although Gemini too changes facial features a little when generating headshots, your product is going a little more off-mark i.e. creating sticker like pictures that fail to match the original features by 30-40%. What I would suggest is providing a feature to prompt the tool to enhance and refine the generated image because in the current workflow, user's are bound to just accept the result without any liberty to tune their headshots.