No workstation update yet (or for the near future). I’m in the middle of a move and need to take care of some other stuff that will keep me pretty busy. Midjourney is my creative outlet at the moment. Still playing with the versatility of the tool. It is a lot of fun once you realize how certain words and or phrases evoke image results. The dataset has some interesting interpretations of words when used certain ways.
(BTW – I’m considering a lengthy rant on those who are against AI image creation tools. Not that anyone would care what I think but – hay 🙂 )
Some fun things I felt like upscaling.
Battle Bot workshop
_____________________________
Friendly Mushroom family on a stroll in the forest
After many (many) years I have decided to move on from my trusty little Dell desktop that I run 3D Max on to a state-of-the-art system. There are a fair number of tools out there now that I would like to use / explore but require a modern GPU to use. So I am looking at a minimum of 16GB system memory / 8GB video memory.
I’ll consolidate from my antiquated desktop and an aging laptop to the one system.
This means no more 3D Max. It’s too expensive to update (along with the laptop cost) and also maintain license. Adobe is too shortsighted to offer a reasonable hobby version. I already use Wings3D for modelling (free) and will likely switch to Photoshop free alternatives and Blender.
I hate change.
I’ve also become very enamored with AI text to image tools. I’ve played with all the ‘trial’ versions using some variation of Stable Diffusion and Midjourney. I found Midjourney to be the best overall for me and have taken the plunge and subscribed after quickly burning through my ‘beta trial’.
So, for the foreseeable future, expect to see more AI work reflected here. I hope to be able to return to modelling sometime early next year.
Meanwhile, here are a few of my Midjourney pieces. I try not to just speed generate to get something that looks good. I spend a lot of time considering and crafting prompts to get images that are fun or thoughtful. I also tend to do some level of post (in Photoshop) on a fair number to clean and or improve the image. The three below are all very different approaches – Sniper had a lot of Photoshop work done on it, the Leaf Dragon had a small amount and Spores had none.
I thought I would never finish this. The Atolm YTJ6600 Star Wars ship is complete. I got real stuck on the greeble work and then texture application was a beast. I did not think this through enough so that I could unwrap the model. So, since all I really do is a single beauty shot, I decided to ‘hand paint’ some of the details.
I rendered the entire thing out with the main shadow and fill lights is one pass (4.65 hours at 4K resolution) and then the specular lighting layer. I also saved as a PNG (ultra high) so that I would get a crisp Alpha layer.
Then, finally, in Photoshop I grunged up and streaked each panel one by one. Once I got going it was remarkably fast (~5 hours). Yep, not likely to do that again.
I should note, my computer is very, very old. My ‘GPU’ is first generation and only 128MB. Was awesome in the day but….
I knew that building a Star Wars inspired design would require, at some point, the addition of the infamous Geeble! You know, those little nonsensical items that populate otherwise smooth surfaces on spaceships. Luckily for me, the Atolm design was not very heavy on such details so I was spared (or so I thought) the need to overdo it.
I first broke out the area that needed the detailing and moved them to 3D Max. I have a greeble plugin which I applied to these panels. After an hour or so of playing with the panels I realized that I could not get scales and randomness correct for a Star Wars ship.
So.
I concluded that the panels needed to be broken up so that the plugin could be more random. I could then get scale and variation more aligned to what one would expect to see. So, a couple of hours later… Nope, fail.
So.
I opened Wings 3D and copied the panels out into a new project and started slapping boxes on the surfaces and doodling details. I did not want to go crazy because you really won’t see these up close. This, however, took many hours. More than my attention span can handle to be honest. I am happier though with the finished project (greebling). I still need to add small random surface details to the hull which will be fairly simple.
Next big project is the engine exhaust detailing and then (dreaded) smoothing checks.
Just finished a few weeks off with family visitors, very exciting and very busy! So, not much (rightfully) going on with hobbies otherwise.
I’ve continued to play with AI art text-to-image generation. I think I’ve now tried (trying) 6 different tools. Playing with all the ‘free’ ones right now to experiment with prompt writing. It is also interesting to see how certain prompts transfer to other platforms. I plan, soon I think, to sign up for Midjourney. Midjourney seems best but you burn through your plan very quickly. So, for now, I am mostly practicing with variouis Diffusion colabs and NightCafe.
The Google Diffusion colab, which is pretty awesome, takes a long time to render (the free version only uses low end GPU’s if you are not familiar which drags performance). While I run batches I have had some time to continue with my Star Wars (Atolm) project.
Lots of panel cutting and extruding. And, as with first attempts noted previously, I’ll render progress shots to then find errors in what I did. These WIP images are the latest after a few ‘redos’.
I have started the detail work on my Star Wars (Atolm’s) YTJ-6600 ship. I have a new found sense of appreciation and awe for modellers that cut in there panels. It is tedious and can be fraught (if you don’t know what you are doing) with challenges.
I am on my 5th version (cut in). The first four had various issues I found in render tests. Figured out what I did wrong, regrouped and (think) I solved it. So now I am applying to everything this particular model will need. Image below (V5) took about 2 hours to get the cuts and panel extrusions completed.
Big craze now is AI generated art from text prompting. There are a load of tools out there and they are all mostly beta (and have pay walls). I’m using Disco Diffusion in free mode. It has lots of quirks (is a Google Notebook Lab, little to no documentation, time limited, resource limited, ya know – Google).
I’m also playing with and likely going to pay for Midjourney.
Anyway, not going to offer thoughts on or debate the whole ‘is AI art?’ discussion here but thought I’d share a few things.
Dieselpunk Abandon Train – Disco DiffusionDieselpunk Abandon Train – Midjourney Hobbit Home – Midjourney
Fun stuff and can be done while I’m doing other modeling stuff.
Yes, a Star Wars related project. The last time I did a SW project I had hair on my head. Yeah, that tells you nothing. Suffice it to say, I only did a video using someone else’s mesh for my company (at the time). We had a COMDEX display and had all kinds of 3D software created content streaming and mine was a X-Wing zipping by with a pretty decent (for the time) planet in the background. Another of my small bits of 15 minutes chewed up.
But…. I digress.
Many of you know of Atolm’s various Trek designs. He also does many other works themed from other shows or genres. I’m taking a crack at one of his recent SW designs. I took the perspective drawing and printed it and then dusted off my old drafting tricks to get the dimensions.
It looks close but I may need to practice those tricks some.
This is a couple of hours work with a ton more to do. Studying SW ship details a lot to finish.
I'm a work'n on it 