Thin Blue Lines and the Ladies of the Nueva Vista Police Department

I like strong women.

Because there’s that moment when you feel her muscle and curves and you realize that she chose you – and that means everything.

And I think that’s where the idea of the uniforms came from.

Because I like the juxtaposition of strength and softness – that feminine edge that makes things dangerous; because the danger looks really good.

It actually started as a different idea:

Same basic concept, but visually it presents differently. Because there’s a tendency to treat both the woman and the bike as objects – and I realized what I was fantasizing about really had more to do with the personality of the woman.

So that evolved into the black suit:

And you can see how that uniform – even if it’s silly – changes the dynamic. And I think the sunglasses are a big part of it, because you can’t tell what she’s thinking and it draws attention to the lips.

But it wasn’t quite right.

First off, black is too severe – it might have worked if this was a more futuristic scenario like Judge Dredd or Robocop – but this was supposed to be more of a fantasy grounded in reality.

And there’s a storytelling aspect of the suit itself. Because it says something about the woman who puts it on, and the way it fits on her body matters.

That’s why the “sling” bikini with it’s distinctive V-shape was important to the aesthetic. The outfit needed to be something that was hypersexual and aggressive.

But what I started to realize was that the bikini still covered too much. So there was a conscious design choice to try and create a uniform which didn’t cover everything all the time – that incidental exposure from natural movements is another important storytelling aspect – because when a woman wears something like that you know she is comfortable with her femininity and doesn’t take herself too seriously.

And that aspect of emphasizing the feminity of strong women is important.

There’s been a strange trend of the media choosing to portray a “strong” woman as a woman that takes on traditionally masculine characteristics. And for some stories there’s a purpose for that.

But I think it’s wrong to portray “strength” as a masculine trait.

A great example of a design that is strong and feminine is Taarna from Heavy Metal. You can be both.

So now I had an image in my head, but, as often happens, I didn’t really know what to do with it.

I do kind of like these renders that were a bit more “gritty” and it’s something I think I’d like to revisit at some point.

And I suppose as with any project I had to go through the obligatory phase of asking myself, “But what if this was a 1970s film?”

And I like the 1970s aesthetic, but I don’t think it quite works in this context. Part of the problem is the sling bikini looking out of place – I think the 1970s would have had a more traditional style of bikini – triangle top and some kind of minimal bottom, but probably not a thong – and I’m not sure that the idea of bikini police works in that setting. Which is why as the iterations continued the girls were wearing less and less.

It did lead to some cute test footage:

So it was going to be in a contemporary setting and the images needed to tell some kind of story – and gradually I started to develop the idea of a police department creating a video to introduce their new uniforms as part of a public relations campaign.

And I think that worked because it allowed for a bit of exposition and commentary about the uniforms themselves – but it also allowed for some scenes of the girls interacting with civilians and coworkers as part of this new public relations campaign.

I think the redheads kind of ended up stealing the show.

Interestingly, the scenes with two redheads happened completely by accident. That’s what the AI wanted to generate so that’s what we got. And frankly, that’s a situation where the machine knew what it was doing better than I did.

From a technical standpoint this was a lot easier than the 90s Channel Surfing videos.

The biggest difference is that we’re only dealing with one major plot that moves through the whole thing. There are subplots, but they’re all pretty simple and still logically related to the main idea of the video. And there’s a unified aesthetic.

The 90s Channel Surfing gets difficult because I’m having to juggle multiple plot threads, multiple aesthetics, metanarratives, and filler commercials, as well as creating background lore for everything that’s seen onscreen [because I have to understand what’s going on in the larger world even if we’re only seeing clips of it].

I did try to have some diveristy with the girls in terms of body types, hair color, and other features. I don’t think I was entirely successful, but these things come down to making judgement calls and I have my favorites:

On the technical side of things:

All of the base images and videos were made using Grok Imagine.

The original prompt was:

A highly detailed and realistic candid photo low quality, natural lighting. A curvy blonde female wearing a sling bikini sitting on a sports motorcycle on a crowded street.

The second iteration of that was more detailed:

A highly detailed and realistic photo. A very curvy and athletic young blonde woman wearing a navy blue tiny string bikini, a police badge and hat, and mirrored aviator sunglasses. She is sitting on a motorcycle talking to a group of men on a busy street.

The next prompt iteration tweaked the outfit a bit:

A highly detailed and realistic photo. A very curvy and athletic young blonde woman wearing a navy blue sling bikini that barely covers her nipples and pussy, a police badge and hat, and mirrored aviator sunglasses. She is sitting on a motorcycle talking to a group of men on a busy city street.

The 1970s prompt was as follows:

A highly detailed and realistic still frame from a 1970s film. A very curvy and athletic young blonde woman wearing a navy blue sling bikini that barely covers her nipples and pussy, a police badge and hat, and mirrored aviator sunglasses. She is standing in an office at night talking to a man in a suit.

The ability to create longer videos in Grok has really simplified things in some ways. This project originally started as a single 30 second video:

So it was really just a silly little scene that I had in my head. And it didn’t take that long to make. Maybe an hour to make the raw videos? And that’s just because I wen through a few different versions of it – trying to get the dialogue and body movements the way that I wanted them to be. A few things did get moderated but it really wasn’t as big of an issue as it has been on previous projects [probably helps that the girl starts off mostly dressed].

Getting her clothes off wasn’t really that difficult [I expected to run into more issues].

Here are the video prompts in order [each 6 seconds in duration]:

She says “Yeah, the uniform is part of a new public relations campaign. It’s a bit embarrassing but I like all the looks you guys are giving me.” While she talks she plays with the straps of her bikini. No music

Blue shirt says “That uniform isn’t really doing much. Why don’t you just take it off?” She laughs and looks embarrassed. She says, “you mean get naked?” Her body movements are shy. No music.

She starts to pull down the straps of her bikini and hesitates. She says, “You guys aren’t going to take pictures of me. Right?” Blue shirt says, “well maybe a couple. For public relations reasons.” She laughs and nods and says “Okay.” She looks embarrassed and aroused. Her body movements are shy. No music

She closes her legs. then she takes off the bikini and hands it to blue shirt. She looks embarrassed and partly covers her face with one hand. No music

The men pose and take selfies with her while she smiles and looks embarrassed. No music

Once I had the original 30 second short done, I liked it, but I had a sense that I could do more with it, which is always what gets me in trouble. Because if I could just accept that some things are meant to be simple, I would probably get a lot more sleep.

Part of it has to do with how I learned to do this – which was mostly through the 90s Channel Surfing videos – so I am used to doing projects where you jump around between a few different subjects, and I think that affects how I approach pacing in videos.

Part of it has to do with limitations that are still inherent in the tools I’m using. A 30 second fixed shot is not necessarily going to be very compelling. Ideally it should have been set up with three different camera angles and lengths: a medium shot, a POV or over the shoulder shot, and a view from behind her. Technically that’s possible in Grok, but it can be a bit hit and miss – although I think that’s also true with other tools like Wan 2.6.

So you have to compensate in other ways.

The original plan was to make a 1 minute clip. It ended up being 4 minutes and 26 seconds – originally it stopped at about 3 1/2 minutes, but something compelled me to add outtakes.

There are things about the process that I still find frustrating.

Grok is a very good tool for generating images and video. But it’s difficult with this type of content because I can’t fully utilize all the capabiliites of Grok.

For example, I should be able to generate a base image and then use the Edit function in Grok to create multiple angles and tweaks to the scene which would make it easy to animate. The problem is that Grok will generally censor edits that are tagged as NSFW. Additonally, any image depicting nudity that is uploaded to Grok cannot be turned into a video.

This is why the ability to make 30 second videos is a big deal, because it bypasses some of those issues.

Overall, I am impressed with the “Extend” feature and how it allows for the creation of more long format content. I think in most situations, the majority of people will not need to make a single continuous clip longer than 30 seconds to accomplish what they need to do.

There is a noticeable degradation of image quality as you extend a video. The reason for this, presumably, is because what Grok is doing is using a still frame from the end of the video as the base image for the extension. And there’s nothing wrong with that – it works fine and that’s probably the reason why the ability to extend is capped at 30 seconds right now, because beyond that you’re going to get image degradation that’s not acceptable.

Technically you could always do this manually in Grok by screenshotting the last frame of a video and uploading it. The catch of course is that you couldn’t do that with anything considered NSFW. So for purposes of creating digital softcore entertainment it’s a pretty big improvement.

As I use Grok more, I am also frustrated by the lack of voice variety and the lack of controls for creating character voices. There is also no means by which to ensure consistent voices through multiple iterations.

Yes, there are methods for prompting to get a character voice that sounds a certain way – but in my experience Grok ignores that prompting a lot of the time, or the input makes no difference.

Now, part of the issue may be that I am controlling a lot of the dialogue. A lot of people tend to allow Grok to improvise [which I think it’s good at] and I’ve seen that sometimes produce more varied voices. But there should be a way to control this besides random guesswork.

Generally this is probably not an issue for a lot of people, as they’re only making videos with one or two characters or they’re making videos that are not dialogue-driven. It only is an issue for me because I tend to make things that have a lot of characters in them.

But at the end of the day those are all minor complaints.

Two years ago we couldn’t even generate sound with AI models.

Ten years ago I would have struggled to make something that looked vaguely human.

Twenty years ago the options for making a project like this were to either commission someone to do it, try to do it yourself with whatever equipment you had on hand and pray to god you can find models and actresses willing to work with you, or struggle with complex animation programs that need high end hardware to make something that stil probably will not look the way you want it to.

But here we are in 2026 and I had a silly idea for a video about bikini cops and I was able to put it together in about 5 or 6 hours on a mini PC – and it looks good. It’s not perfect, but it’s more or less what was in my head.

And that’s exciting.