1. Introduction and disclaimer
This tutorial aims to show how to generate interesting AI imagery in a fast and thrilling way.
Leonardo has just unveiled their new "Flow State" mode, and this helps a lot, too, and might be a real game changer for AI art!
I used to create a little "music video" that runs for close to 3 minutes (okay, maybe a few seconds less ;-) in *under the time* of the video!
Try to beat this with "traditional" methods of video making.
Note: I used a movie technique called "montage" for the video. This was, and is still considered a high form of art. While being ubiquitous in movies and media, in more mainstream type of movies, "montage" sequences are usually delegated to a lesser role, such as openings, dreams, moments of romance, travel sequences... (remember Harrison Ford in "Raiders of the Lost Ark"?)
montages often lack traditional narratives, or structures, and can be dream-like, "stream of consciousness" cuts.
Thus, some viewers might find the video at the core of this tutorial to be a bit unusual or strange, but that's intentional.
Also it's aimed to portray a futuristic / scifi / parallel dimension "vibe" which might not be everyone's taste either. or to use more colloquial words: "the sounds and visuals are trippy as hell, dude!"
Despite of this: I'm not a top notch video producer, it's meant to be rough and gritty, and the visuals were done in under 3 minutes.
So I'm certain that someone with skill, more time and patience could create something for better and more stunning :-)
Alas, it's a tutorial to built on - for you!
And this gets us to the next point:
This is not some "one prompt fix" automated AI video generation where you can sit back and relax.
It's meant for creative people and to show how AI can *help* with creative projects - not to replace it!
Not to replace your work, not to replace you, and not to feed your possible laziness :-)
So this not for people who want to go to some online AI tool, type in "generate video of Godzilla eating a hotdog" and voilà, it's finished.
Anyway - let's get to work now.
2. The tutorial
For this tutorial, I wanted to:
a) create images that could be used in a video to accompany a futuristic electronic music track ("Metalove - Maia (Ich Will Sehen)")
My workflow is this:
1. I ask ChatGPT for ideas and prompts for the image generation
As it should be a futuristic video, I decide on the three "seed" ideas:
1. futuristic / impossible cities
2. cyberpunk world
3. cosmic rulers and imperatrixes
this is some "human brain" work that i did - i could have taken a short cut though, and directly asked ChatGPT "show me ideas for a futuristic video with ai visuals"
either way.
2. now the "work in 3 minutes" segment starts.
a) i go to chatgpt and ask it to brainstorm prompts for me to use with leonardo.ai
b) i quickly choose 5 prompts out of ~40 that i like best and copy them into a separate text and give them the numbers 1-5
c) i go to leonardo.ai and open flow state
d) i paste the first prompt into flow state and click generate
in its current state, flow state generates 16 different images of the most varied and diverse kind.
it explicitly does not strictly adhere to your prompt - it gives you output that ranges from pencil sketches to renaissance paintings, from manga art to scifi illustrations, from pixel art to Art Nouveau.
and from a single prompt, you can always create more.
thus - there is a lot to choose from. coming your way, generated quickly. ideal for projects where you need to have a high yield of creative output.
e) i quickly choose around 12 images which i like from the yield
f) repeat the same run with prompts 2-5
g) now i have around 60 images to use for the video illustration!
h) i load them into my video program and edit them according to the abovementioned "montage" technique.
voilà! it's finished!
a quick and simple way to do a video project.
and... you know what they say?
i) "the sounds and visuals are trippy as hell, dude!"
So, this was just one way to work together with ChatGPT and Leonardo.
It was a rather simple and easy way - if you put some dedication to it, I'm sure you can build some very impressive works!
Other ways the above method could be used:
1. I could have used leonardo's ai animation tool on the images, and would have had a more traditional "video".
2. Someone could use the images as a "background slideshow" to a real video. envision a human scifi rapper rapping "in front of" a background of various galaxies, planets, and star systems!
3. the images could be used for something entirely differently, like illustrating a story on a blog (in a quick way).
4. instead of creating surreal and space-y images, they could have been more down-to-earth, art-like - or even going in the fantasy, RPG type direction! etc.
and and and...
use your own mind on this. the possibilities are endless!
3. the end
I hope you enjoyed this little tutorial, and that it might be useful for you in some way.
again, if you think it "still" looks a bit cheap - this is just a tutorial!
with some dedication, wholly different and more polished things are possible.
if you have further questions - feel free to reach out to me!
4. addendum
you can watch the finished video here:
https://www.youtube.com/watch?v=q3anZoc1-n8
example for one of the prompts i typed into chatgpt:
"i want to generate surreal ai art with leonardo.ai.
please give me some prompts that depict the most surreal, futuristic, strange new worlds and parallel dimensions"
example of one of the prompts chatgpt brainstormed for me to use with Leonardo ai and flow state:
"A gravity-defying city floating in the void of space, with buildings made of liquid crystal and highways that ripple like water, lit by glowing orbs of plasma in a palette of neon blues and pinks"
More information about the Montage Film technique in art and media:
https://en.wikipedia.org/wiki/Montage_(filmmaking)
No comments:
Post a Comment