>>8664734
>>8664716
AI models work by being trained on many input images and caption files, then when you prompt an AI it's running a predictive algorithm based on what it's learned. So if you want to replicate a specific type of clothing there would need to be enough example images of that clothing in the training data for the AI.

Modern NoobAI based models were trained on a Danbooru and e621 data set. Danbooru has a list of different lingerie at https://danbooru.donmai.us/wiki_pages/lingerie . One of them is the Teddy that you're talking about: https://danbooru.donmai.us/wiki_pages/teddy_%28lingerie%29

However, it sounds like you're looking for a very specific look. There's still a lot randomness involved in AI generation so with enough attempts what you want could randomly generate. Or like >>8664731 said you could train an additional file called a LORA on images of what you want and then using that should work. Lastly, you can always use editing and inpainting to tweak and AI gen to be exactly what you want, so you could take an image that's close and edit it from there.

Here's Dejiko showing off all the different lingerie tags from Danbooru. Teddy is the last one though the pose doesn't show if off much. If you're still around in the thread I can try some more attempts at it to get exactly what you're looking for.

Individuals at: https://imgbox.com/g/dbmwNaRHmg