I think to get higher resolution, you’ll have to use a separate tool altogether. Tons of them are popping up nowadays.
And I hear you on the age problem. It’s hard to get characters who aren’t in their twenties but also aren’t old.
I spent weeks trying to create Ian’s mom. In the end, it was worth it but it sure was frustrating.
I’ve given Dumbledore a try again and these might be my favorite. lol
They don’t match the style of the other portraits in this series but they’re so… fabulous.
Okay, let me try to evolve the style to resemble the others.
The closest so far.
Maybe that last one would do but I was hoping for a longer beard…
If I could extend the beard, then maybe this one? It has a nice fabulous vibe going for it.
I’ll experiment with some beard evolves later.
In other news, I’ve been compiling the characters of this series into one massive collage. It’s looking great so far.
There are a few adults and a few side characters to add. Maybe I’ll put them on a second collage or maybe I’ll make a version with everyone.
The attempt to create the characters of this series has been a lot of fun so far.
Yup! Honestly, I am glad the AI is here.
Especially since it doesnt really need any serious edit that I can see. It’s rare to get one that’s good enough, on some filters.
That’s amazing. Good job.
I really love this filter. It’s perfect for romances!
There you go, Rivals and Revels has new cover! Effing amaaaazing!
If you get enough of them, you got some headers that suit the book.
Yeah, I just keep it churning through while I work. It’s great. The resulting covers are nothing like starting images whatsoever.
Like this transformation:
From this:
To this:
I like it. Are you in love with orange? Because that’s the easiest thing to learn to change in an editing peogram. Orange is eye-catching, though.
Maybe I canceled my Mid Journey subscription too soon.
Apparently there’s also a new variant of Stable Diffusion being developed.
Hoping for a local install option for both.
Somewhat related. Just found these videos about text-generating AI software (local install). Could be useful for character dialogue trails / inspirations.
Late update: I installed the Oobabooga AI-software* and experimented with a few chat models and character conversations (default settings). Not impressed (really short chat responses). Perhaps image generation is easier than authentic, random dialogue. May need to wait a few months for narrative AI software to catch up to MJ and SD. Alas, Halo style AI assistants will remain fiction for a long while yet. Might uninstall / delete and stay with Chat GPT for now.
*My install experience / process was different to the first video’s description. Also SD and MJ seems to have a much larger (online) custom models / loras community, and overall public interest. Text editing and text summarising AI-software is mostly DIY (or not open to the public yet (cough GPT4 cough)).
I’ve been looking around for animation tools.
This one is meant for animating a full person.
Be aware, the terms are very sketchy (lol, that’s the name of this page). Anything you upload you give to Meta. At least they’re honest about it.
I used a HeroForge version of Ruby and created this funny little thing.
Clearly, it’s just a toy. I hope in the future they’ll come up with something more serious.
I kept looking and found more:
This one adds depth to an image. It’s actually very easy to use and pretty amazing how well it works. It does one thing and it does it well.
Here’s Jack.
There are a lot of options to customize your animation but one thing I did not figure out is if you can just have a nice and long zoom in or out instead of the swinging motion.
Myheritage site has a bunch of AI toys now too. You can animate a portrait to add facial expressions or even to make the photo speak. As I’ve found out, there’s a limit how many you get for free before they ask you to subscribe to their 14-day trial.
Here’s Colin.
It’s cool but I don’t like how they animate the mouth. It’s a shame. There’s potential here.
In this one, you upload multiple photos and I’m not sure what the result is, but it sounds a bit like Dreambooth. Hmmm. Curious.
And it’s 4am. Sigh. At least I got to hear the morning robin sing.
There is a WebUI extension option for Stable Diffusion (local install) that can turn 2D images into 3D imitations (I think). I’ll research and experiment with that tomorrow. It’s a tad late here too.
I have played with Reface for some animation.
Experimented with something a bit more complex. Lots of failed images. Here are some of the successes.
NightCafe has a new model SDXL 0.9 (who comes up with these names?).
Very expensive (6 credits each) and slow because it runs longer?
Initial impression: works best when the prompt is short. If you give it too much info, it messes it up, but if you’re exact, it gives quality result and high resolution.
Lamassu has been difficult to create in the past. Believe it or not, it’s hard to create a winged cat. It defaults to creating humans in cat costumes so this is a nice achievement right here.
From my past experiences trying to create her, I noticed that AI can’t handle the term “black panther.” It just defaults to a superhero costume. So I started using “black lioness” and that works better. it actually gives me cats.
why boobs? Otherwise, nice quality.
This would be very nice if quality was better.
I submitted this one to today’s challenge (cats with jobs) because why not.
Hmmm. Actually, the last one resembles the HeroForge version of Lamassu. Let me dig it up.
There she is. Look at that face!
Huh. Is that cool or what?
Btw, HeroForge is not AI. Takes some work but it’s awesome for fantasy races. I’ve used the characters created in it to evolve in Dream too.
Since this new model is so expensive, I stopped digging in to find a better prompt. I’ll likely resume once they bring down the cost a bit. There’s still room to improve her.
I gave her gold armor because it looks really cool against black, but now I’m curious what it will do if I don’t mention the armor. Okay, 6 credits, let’s go.
omg, it’s running forever!
You know, I was looking back at your car pics, thinking, wow, these are cool cars. Why doesn’t my story have cool cars? And then I remembered the cars I’ve created for my story…
I’m so not cool.
hmmm. I didn’t create Melody’s car yet. Okay, here we go.
Interesting. The car description greatly affects the setting description (the only setting prompt I gave it was “parked in an alley”).
When I didn’t call the car old, it gave me a fairly normal setting
Junky car = junky neighborhood
Old car but no junky descriptions = uncared for but normal setting.
I was using SDXL beta for this which is 3 credits for 4 images. I sure am not spending 6 credits on an old clunker. lol
I haven’t been able to create the mismatched door. Maybe that would have to be edited in Krita.
Anyway, so the cars in my stories suck.
On the other hand, the other story has hoverpods instead, and those are cool.
Update: omg, I know what it did! It took my car door descriptions and just inserted a bunch of doors in the background. That’s why it didn’t give me the mismatched door I asked for!
