I Dove Back into AI and Returned to Windows

The one downside of leaving Windows for Mac is being unable to render AI images locally – at least not with the degree of flexibility available on Windows. Most apps and models are Windows-only. The only Mac app I’m familiar with is called Diffusion Bee, but support for it is lackluster and its UI is unintuitive. Consequently, I’ve had to relegate myself to web-based services. Unfortunately, the best AI model I’ve found isn’t free or NSFW, so I’ve had to be extra-creative and think outside the box to come up with text prompts that won’t return ToS alerts. Nevertheless, it will not generate nudity, and provocative clothing is a no-no.

The model I’m referring to is Google’s Imagen 4, and it’s available on Krea.ai with a subscription. Not even Flux – a widely popular open-source model – matches its visual fidelity or prompt adherence.

Thankfully, a close family friend – we’ve known each other since High School – gifted me a one-year membership to Krea’s Max plan for my birthday last month. Suffice it to say, I went a little nuts, and have lots of new content to share. Since Krea doesn’t allow NSFW content, I wound up returning to Windows so I could use Fooocus to inpaint nudity offline. I love my new iMac, but again, it’s not well-suited for AI, so I re-purposed it for video editing.

In other news, I re-arranged my Galleries. My 3D and AI art now have their own menu structures, and I’ve created separate pages for my AI Enhanced and AI Generated work.

With all of that in mind, one thing I discovered about Imagen 4 is that it knows who Michael Myers is. He’s my favorite Horror movie villain, and I created a new gallery to showcase his exploits with ten images to start.

There are seven new one-offs on my AI Generated gallery: They’re called Strangled Attorney, Impossible Demands, Strangled Streamer, Cop Killer, Behind Glass, Case Dismissed, and Hitman’s Prerogative. I also created my first – and only – multi-panel AI comic, called Whistleblower.

Finally, I experimented with the new Wan 2.5 text-to-video model. You can read about that and watch my test clips on my new AI Videos page.