I had been using Flyne.ai as a way to access nano banana 2 and VEO 3.1 as image and video creation models. I could do full nudity and probably even some X rated stuff, but then yesterday any prompt I would type would come back with "Restricted Content" warnings, worse than the early days of Bing. So Flyne has decided to implement this sudden censorship, where before that, they were said to be 'NSFW Friendly'. Looks like that has come to an end, and I have credits I can't even use now.

I've posted an actual screen shot of my innocent image and prompt. I had taken terms like 'in the face' and 'cream pie' out of the prompt, but they still won't accept it. I chose them because this was not an issue. . . till yesterday. So this fun little video clip was not created as a result.

So I'm in search of 'NSFW-Friendly' image and video creation options. Any suggestions? I like using Nano Banana 2 and I don't think it's them that created this problem, it's Flyne that I was forced to go through.

I mentioned it in the past but I have strong doubts that Flyne was outputting using Nanobanana 2 or Veo3 for the full nudity and X rated stuff. I suspect it was rolling over to an open model. I've tried numerous NB direct api's and not one allows creates nudity even with guardrails down. The images you posted didn't have a synthID which is mandatory encoded on all NB2 and NBPro images.
AS far as imaging goes, I doubt you'll find any NB provider that will create what you are looking for. For video, you're looking at Grok1.5, Kling3 or Seedance2. Seedance is by far the best quality and prompt understanding but even then you need to choose a provider that doesn't place their own guardrails ontop and also use one of the less restricted models. There are a number of seedance2 model versions available such as the chinese Volcengine or a Bytedance variant. Different versions will have different restrictions on Uploaded faces, IP restrictions and output content.

I've been using Go Enhance. It seems alright. You get access to several AI generators.

I feel your pain. I have been using Reve AI for wetlook. It 100% knew what straight-from-the-water clothes and hair should look like. Yeah, only about 1-in-4 was good, the others would have weird morphing or other issues, but the good ones made it worth it. 2 days ago they rolled out the newer, better, updated Reve 2.0. Yeah, the images are cleaner and more crisp, but it lost 75% of it's wetlookability. If you are looking for got-out-of-the-pool-10-minutes-ago wetlook, then its probably amazing. Output file size is 20x larger too. I was in the middle of SEVERAL projects, trying to get prompts just right or waiting for the perfect image and so close, now it will never happen for any of them. And I was just starting to get over the loss of Grok over their new moderation and limits... yeesh...

Edit: Maybe I need to give it time. Mess is way better than the previous version, as long as it isn't a "wet mess" that would act the same as water. It is also way worse at reference adherence which it was pretty good at. It can't even get the aspect ratio right half the time right now. It might be overwhelmed cause there are a whole lot of people just hearing of it for the first time, half baked because it just got uploaded, etc.