I never wanted this.
I never will.
That is where SHAKEN started.
The song is about what happens when the past gets into your nervous system and refuses to leave. The body remembers. The mind strategizes. Something underneath you keeps fighting while another part of you is saying, “Seriously? We are doing this again?”
Then comes the line that became the whole point of the video:
I’m shaken
But I’m standing
I originally imagined SHAKEN as a live-action, cinematic 1990s alternative-rock video. Storms. Empty rooms. Cracked concrete. A woman walking through increasingly hostile landscapes.
Then I looked at what I was making and thought:
Why am I trying to make this look real?
Maybe it shouldn’t.
2/9
THE FIRST PROBLEM: TRAUMA IS NOT VERY PHOTOGENIC
Well, actually, it is.
Hollywood has been making trauma photogenic for decades.
But I didn’t want guns, blood, screaming faces, or somebody dramatically collapsing in the rain. Those images tell the audience what to feel.
I wanted something stranger.
I wanted the inside of the experience.
So I started building visual metaphors.
Cracks spreading beneath someone’s feet.
A tunnel disappearing into darkness.
Hands gripping rope.
A storm trying to knock someone over.
A black pit that might be dangerous and might simply be familiar.
A house in the distance.
A person who keeps walking.
Lab note:
The more literal I became, the less interesting the images got.
The more metaphorical I became, the more the song started talking back to me.
3/9
BUILDING THE VISUAL LANGUAGE
I broke the song into emotional sections rather than simply illustrating every lyric.
- The shaking begins
The early images are cramped and restrained. Abandoned rooms. Concrete. Dust. A body standing still while everything around it trembles.
- The chorus explodes
The world gets bigger.
Prairie. Wind. Cracked earth. Storm clouds. The woman is still there, but now the environment is practically attacking her.
- Memory becomes visible
The story moves into an old house.
Photographs.
Thread.
Objects that have been broken and repaired.
Beautiful colors that suddenly become disturbing.
- The pit
The bridge gets claustrophobic.
There is a black opening in the floor. Something is moving down there, but I deliberately didn’t define what it is.
Because sometimes defining the monster ruins the metaphor.
- Home
The final images aren’t about victory.
They’re about movement.
A small house.
Morning light.
A woman standing there.
Still here.
4/9
THEN I CHANGED THE ENTIRE MEDIUM
This is where the fun started.
I tried pushing the imagery toward a rough charcoal-and-ink illustration style.
Suddenly the imperfections became useful.
Charcoal smears could become shadows.
A rough line could become a crack.
A patch of black could be a doorway, a pit, a memory, or absolutely nothing.
And color became something I could use almost like an instrument.
The verses could live mostly in charcoal, gray, brown and faded blue.
Then the chorus could erupt with color.
Not pretty color.
Emotional color.
The visual world could literally become more unstable when the music became more unstable.
Lab note:
This is one of the things I love about AI image generation.
The tool doesn’t merely execute the concept.
Sometimes changing the medium changes the concept.
I started out asking the machine to illustrate a music video.
I ended up discovering what the music video actually wanted to look like.
5/9
THE 30-IMAGE PROBLEM
I decided to build the whole thing around 30 major images.
That sounds reasonable.
It is not.
Thirty images is enough to tell a story, but not enough to hide behind endless coverage.
Every image has to earn its existence.
I needed continuity in:
- Character
- Clothing
- Lighting
- Emotional progression
- Camera language
- Color
- Environment
- Symbolism
And then there is the usual Midjourney problem.
You ask for one thing.
It gives you something else.
Sometimes the “something else” is terrible.
Sometimes it is much better.
Those are the dangerous ones.
Because now you have to decide whether your carefully constructed plan is actually better than the happy accident sitting on the screen.
6/9
THE CHARACTER HAD TO STAY REAL
I eventually settled on a consistent character description:
A Caucasian woman in her early 40s with shoulder-length dark brown hair, weathered expressive features, wearing a faded charcoal jacket and dark jeans.
I repeated that information constantly.
Because AI has the memory of a goldfish with a very expensive camera.
If I told it “the same woman,” that wasn’t enough.
So every prompt became its own little universe.
That meant repeating the character.
Repeating the visual language.
Repeating the cinematic vocabulary.
It is not elegant.
It is effective.
And somewhere along the way I stopped thinking of these as individual pictures and started thinking of them as frames from one strange illustrated dream.
7/9
THE TOOLBOX
The main visual stack for this project was wonderfully simple:
Midjourney
–sref 5326
–v 8.2
–seed 777
–s 250
–c 10
The style reference became particularly important.
It pushed the images away from generic photorealism and toward something much more tactile and illustrated.
I wanted the pictures to feel like they had been dragged through charcoal, graphite, old paper and a slightly unhealthy amount of emotional weather.
The music itself was built around a 1990s alternative-rock vocabulary — stripped-down verses, then a much harder chorus with distorted guitars and aggressive drums.
The visual arrangement followed the music.
Small.
Tense.
Explosive.
Quiet.
Explosive again.
Then almost nothing.
8/9
THE HAPPY ACCIDENT
The biggest discovery wasn’t an individual image.
It was the realization that SHAKEN doesn’t need to end with triumph.
I had initially been thinking in conventional music-video terms.
Character struggles.
Character confronts darkness.
Character wins.
Roll credits.
But that isn’t what the song says.
The final lyric is:
But I’m still here.
That’s much more interesting.
So the final image became deliberately ordinary.
No heroic pose.
No giant sunrise.
No arms raised toward heaven.
Just a woman standing inside an ordinary house in soft morning light.
No smile.
No tears.
Just presence.
That image suddenly made the whole video make more sense.
The point isn’t that the shaking stops.
The point is that it doesn’t get the final word.
Lab note:
Sometimes the strongest ending is not “I conquered this.”
Sometimes it’s “I’m still here.”
9/9
THE REAL LESSON
AI is very good at making pictures.
That’s not the interesting part anymore.
The interesting part is deciding what the pictures mean.
I can type a prompt and get a beautiful storm.
Big deal.
The real work is asking why there is a storm.
Why is the woman standing in it?
What happened before this?
Where is she going?
What does the color mean?
Why is there a crack beneath her feet?
Why does the house matter?
Why does the final frame need to be quiet?
That’s where I become the video alchemist rather than the guy pushing buttons on a website.
The machine supplies possibilities.
I supply the obsession.
And somewhere in that back-and-forth, something neither of us completely planned starts to appear.
That’s my favorite part.
TL;DR:
I started with a song about being shaken.
I tried to make a realistic music video.
Then I realized the song wanted charcoal, shadows, fractured landscapes, violent color and a woman who refuses to disappear.
So I followed it.
Sometimes the best thing an AI can do is misunderstand your first idea well enough to help you find the better one.
And after all that shaking, breaking, breathing, remembering and walking…
I’m still here.
Steve Teare
video alchemist
TerminallyBored.Monster
Palouse, Washington USA