IMDb does allow porn, you need to go through advanced search and enable it to find it but you can find it if you are looking. There is also more info that is hidden from the profile unless you have an industry account. I believe they allow it for this reason as there is no reason why a make up artist or camera person can start in porn and move to mainstream.
I would recommend you download files to a nas and use xbvr or stash + stash-vr-companion (a tool I wrote) or stash-vr and stream content to heresphere or deovr running on your headset.
Another of my side projects is timestamp.trade to share markers for xbvr or stash so you get the timestamps of each action crowd sourced from other users.
I am sure you are aware but eroscripts do have people make scripts for videos many are free so join the forum and look for the Google doc of free scripts.
The original version of this scene https://www.sexlikereal.com/scenes/time-stops-japanese-lolitas-19992 There is a group of people going after porn companies to try and cut them off from visa and MasterCard and trying to pressure companies to censor porn. The original version of the scene contained “stop time” themes and was removed from the site. An edited version was put back a year later with the story cut out and just the sex remains with huge edit jumps so a 1:22:20 scene became 38:13. There has been a fair bit of porn purged from the internet and censored a few years ago.
If you are expecting you to just click go and have 10 pages of erotic stories then no it does not exist as far as I can see. Ollama is a tool to just download and run a llm on a computer and most models are designed to tell you to go away if you ask for anything like that. Some have been uncensored and do not have guardrails but that is not your problem.
A large language model is really advanced auto correct. We have taken all text we can, done some fancy statistics to work out what words are used neer each other and found what the average most common next word would be. Keep rolling the dice and work out what the next most average word is and repeat. A llm does not know what a word is and is based on converting combinations of words to a number and some fancy statistics to work out how common “th” and "e " are so it can spit out "the " randomly when it thinks it is statistically needed.
Llms cannot keep track of context and the more you use them thr more likely they will forget character names etc