Відео

Thanks. Check out some of my (much shorter!) videos of my latest builds! I created Shakespeare's Globe Theatre and Milan Cathedral. And two wizard towers in the sky :)

hiya. I used Stable Diffusion, specifically Automatic1111 webui with Controlnet line drawing: realistic set to pixel perfect, then prompted it with variations such as "beautiful blonde in a cream lace dress"

unfortunately it isn't in the standalone Fusion Studio. It is now available in Fusion inside of Resolve, which it wasn't at the time I made this video, but I now have a tutorial on that method too, which, frankly, is easier and more versatile. ua-cam.com/video/0mzo2G2t51E/v-deo.html

yup, that's pretty much how I use it. Interestingly, in my experiments it works across different models trained with different data, and still produces the same face. Of course, once you've managed to generate four or five consistent images, you can always feed those into a faceswapper and virtually guarantee success! (there's several for Stable Diffusion, which is what I use).

@@thewighteyeguy I am realizing there are so many options for faces - e.g., roop, faceswaplab, lora, textual inversion, controlnet canny, controlnet ip-faceplus, and now this! We are spoilt for choices (and utterly confused!)

yup, pretty much something like "a 40 year old man, Robert Schuster, ....". To ensure it's a bit more unique, you can always use the 'blend' method - something like "a 40 year old man, [Robert Schuster, Martin Bumblethwaite:0.5]". As I replied to your other comment, though, interestingly, in my experiments it works across different models trained with different data. Almost like it is a little glimpse into how AI views people based on names. I thought it was weird it's even a thing! I hadn't seen anyone else really discuss it.

I see what you mean, it is surprising that it works across different models! But I assume SD 1.5 based models? I wonder if it works across SD 1.5, 2, and XL?@@thewighteyeguy

I'll see what I can do. At the moment the comp is a bit of a bloated mess. I kind of shoehorned 3 other comps I was playing with together and it needs refining and integrating a lot more. I may do a couple of short breakdowns to show the basic methods for the friction based particles and the HDRI 3D, which are the two main parts to what's going on here. If I really get my act together I'll maybe release the comp or fx.

@@thewighteyeguy thnks. but try to make a tutorial if you can.

of course, the only caveat to that is that links in youtube shorts comments aren't links, so you can't click that. *sigh*

thanks! It's a tweak of the technique I used for my 'cloaked alien' comp - ua-cam.com/video/AX9wh8X4mNU/v-deo.html and reusing assets from my 'vampire vfx' composite ua-cam.com/video/JSL0jNR-c8k/v-deo.html. :)

hiya. Sorry for the late reply! UA-cam saw fit to not bother notifying me. grr. I can certainly do that! It'd be quite specifically aimed at users of Stable Diffusion and even more specifically at users of the Automatic1111 webUI, but I can certainly put a video together for it!

@@thewighteyeguy can you try to explain how? can it cut the time to make artwork?

@@zhuzet sure. You might need to look up other videos if you haven't got these tools installed and stuff, but very briefly, I use Stable Diffusion Automatic1111 with the ControlNet Extension, set to LineArt, Pixel Perfect, with varying prompts on how I want it filled for a particular variation, such as "photographic portrait of beautiful teenage girl with auburn hair and smokey gothic makeup wearing a vibrant coloured crown of flowers and green leaves. extreme details, dynamic lighting, Highly Detailed, Cinematic Lighting, golden hour, backlit, glow, ethereal, sunbeams, light rays, sunset, back lighting, 8k, HD, Hyperrealistic, micro details, dramatic lighting, ray tracing, 8k ultra detailed, photorealistic, 35mm film"

It's a style ig lol maybe she enjoys breathing heavy af while moving around

I imagine because it was a video from a collection shot in 2020 called 'Social Distance' :) www.pexels.com/collections/social-distance-3f9jog0/

@@thewighteyeguy Thank you for explaining!

Juno has a Scratch-like visual programming interface called Vizzy. It's the thing I show in the first 5 seconds of the video and scroll through. Which was my very linear step by step instructions to get me from the launchpad to orbit. I've progressed quite a bit beyond that now! Including a 'deorbit on command' script :)