Robot Pirate Ninja

With the growth of generative AI, we are now moving into a new generation of models and capabilities. Natural language image editing is a powerful technology. By mixing the near infinite combinations of words with a near infinite combination of image editing options, one is left with an immensely powerful tool, limited only by the users imagination and ability to use natural language to describe what they would like to see.

It is my pleasure to introduce to you, Slopify.online.

I have a few more URL’s that will point to the same tool (including my current favorite slopify.studio..login not working there yet), but the idea is the same.

“Slop”.

The backlash against AI generated art in certain communities has become a bit extreme, and understandable. As someone with a number applications utilizing the technology in a production capacity, I’m well aware of this.

I continue to feel the backlash is (often) misplaced, and a bit of a deflection of other emotions that advanced and advancing technology bring about in people. As we march, full speed, toward the Singularity (I have my own definitions and feel we are in the midst of it right now), it us fully understandable for the deep feelings of anxiety towards technology that, increasingly, feels and looks like magic (but we know is just math).

That’s the difference though. And even among those who “know it’s just math” there is a difference between those who comprehend JUST HOW BIG some of these numbers can get, and those whose mathematical minds maxxed out in algebra.

That this is happening under the backdrop of fascism is not to be missed. I cover that on my other website now (escaping dystopia).

This conversation of focused on the technology itself, not the political ramifications of its existence or other users.

So what kind of power are we talking about here? Stuff like this…(note: these are a few examples that are *why I made this tool*. I hope that other people come up with more and much more fun and interesting use cases.)

Note: as someone old enough to have black and white baby photos, this is incredible technology to me. Typical API response times are about 6-8 seconds.

The speed that one is able to use to modify styles is truly impressive.

The biggest reason I created this tool, however, was to support images created for Rogue Vibe. As with many algorithmically (mad-libbed) generated images, we often find ourselves getting 70-80% images, with one (or two) obvious flaws (and 5-10 minuscule ones). We are not going for 100% perfect here, the monster images in the game can be on screen for less than a second or two as the player mows down enemies. We (for this purpose) need images are that are 95%+ and don’t get uncanny until studied for 10+ seconds, if ever. Bridging that gap can be challenging. Traditional photo editing offers solutions, but at the cost of time and energy.

Recall, humans consume energy to exist and do work. When discussing the energy consumption needs for various generative AI applications, it does no one any favors to do so in a vacuum. Comparable human efforts *when it comes to energy consumption needs* so outweigh those of generative tools it quickly becomes a question of power tools vs hand tool. There is no question that for industrial level work, one needs industrial level tools. Tools like Photoshop.

The cost of *putting out requests* for 80,000+ character images is beyond the cost of generating prototype images themselves via GenAI. The cost of supporting those humans while they do that works is orders of magnitude higher.

Photoshop itself is currently incorporating the model that slopify.online is now using (live in production). They are testing it in beta, and charge premium prices. They have to. There’s a lot of humans they have to support. We (I) don’t have that issue, and can offer a competing product at *much closer to the wholesale price*.

Such are the realities of the world *today*, much less in the future.

A few more quick examples of the power here…(the main one is *character consistency*). Observe:

We start with an image generated for Rogue Vibe

And we make it better, “Add a background appropriate for a AAA dungeon extraction game.” And a few seconds later.

This applies to various setting. Starting from here:

We go Christmas.

We go scary.

Then less scary.

We can go Halloween.

More Pirate and on a boat.

And off to the High Seas.

All avialable in a saved history folder via our new interface at slopify.online

All done via natural language prompts, and the entire workflow completed in a few minutes (as I discussed the next step w/my son). Responses are usually about 5-7 seconds.

Thanks for reading all the way down here. I like to give bonus content to folks to who make it to the end, so this the example of what this tool is called what it is, and why I created it (to fill a need for me).

Lots of AI generated images are “sloppy”. They include elements that no human artist would include. This tool allows one to quickly remedy 90% of that. So if you have a 70% image, 90% of the missing 30%, give you another 27%, taking one up to the high 90%’s in image quality. When needing an image that has on-screen life of a few seconds (in most cases), that’s “good enough”.

So we can take this image..

And turn it into this image.

With natural language, an image credit, and a few seconds.

And some trial and error as folks learn the tools.

And now…we dance.

Enjoy! slopify.online