How to create real looking UGC videos using AI?

I want to create real, convincing and believable UGC videos with real skin texture, real lighting, real body movements, real talking style in sync, and real face expression.
I created my AI influencer image in Nano Banana Pro.

I tried

• Veo 3 : Very polished and artificial and the expressions doesn't match the prompt.
• Sora 2 and Sora 2 API : Sora 2 most of the times keeps showing "Unable to Generate. We're under heavy load". And when it works then it shows error "Suggestive content" for prompt even though it's simply mirror selfie video of a fashion influencer. Same error when used Sora 2 API on 3rd party platforms.
• Kling AI : Worked decently well for video-to-video motion control but it doesn't make good content with prompt. And even in motion control movements are not very smooth and skin texture and lighting seems artificial.

In my opinion Sora 2 output looks the best which I've seen on social media. How do they create content like that? How do I bypass this?

Example


I can't use those AI UGC platforms where they provide their own avatar. I want to use my AI