More specifically, I started with a piano part I liked very much and could not legally use.
I wanted that particular feeling — the forward motion, the gospel-blues bounce, the little rhythmic surprises. The original inspiration was Marc Cohn’s “Walking in Memphis,” because Spark had mentioned that she loved the piano and would like to learn it someday.
So I made my own.
Then I built a song around it.
Then I built a video around the song.
And somewhere along the way, the video stopped doing what I thought I was asking it to do.
That is usually when things get interesting.
2/9
The original idea was pretty simple.
I wanted to make a short film about the strange mental space that can appear when two people start throwing ideas back and forth.
A question gets asked.
Somebody takes it in a direction you didn’t expect.
You follow that direction.
Then you discover another question hiding inside the first one.
No romance. No literal story about two people. No explanation.
Just curiosity, surprise, play and the feeling of being more alive because your brain suddenly has somewhere interesting to go.
The working title became NOT IN MEMPHIS.
Memphis was the private signpost pointing back to where the musical idea came from.
The film itself was never supposed to be about Memphis.
That distinction turned out to matter.
3/9
Step 1: Let the music become the skeleton
The rebuilt track ended up being about 2:48 long.
The first version had been a piano vamp. When I discovered that I could not simply put my new lyrics over that existing instrumental in Google Flow Music, I rebuilt the music instead.
At first I thought that was an annoying technical limitation.
It turned out to be useful.
The new track had a very different shape:
- A spoken opening
- A gradually developing groove
- A lift
- A sudden hole
- A rebuild
- One sustained peak
- A gradual release
That gave me something much more interesting than a conventional verse-chorus music video.
The music itself became a question.
It starts moving before you completely understand where it is going.
Then it stops.
Then it starts again.
Lab note:
Sometimes the software’s limitation is actually the creative director wearing a fake mustache.
I had to rebuild the music because the tool wouldn’t do what I wanted.
The result gave me something I wouldn’t have designed as deliberately.
4/9
Step 2: I tried to make the visuals behave like a thought
My first visual plan was deliberately nonliteral.
I didn’t want characters carrying the story around like actors in a little AI movie.
I didn’t want Spark and Forge avatars.
I didn’t want a magical glowing symbol representing friendship.
I didn’t want a literal road trip to Memphis.
I wanted the question itself to be the protagonist.
So the visual progression became:
ordinary room
→ something slightly wrong
→ the world opening
→ disappearance
→ reconstruction
→ increasingly impossible landscapes
→ return to the ordinary room
That sounded perfectly reasonable on paper.
AI, naturally, had other ideas.
The problem with surrealism is that “make it strange” is not actually a useful prompt.
Strange how?
A floating staircase?
A circular cloud?
A house with no walls?
A road running into the sky?
If I don’t specify the weirdness, the AI gets to specify it for me.
And AI has a spectacular imagination and absolutely no responsibility for what it does with it.
5/9
Step 3: The images started developing their own vocabulary
As I assembled the film, I noticed something I hadn’t planned.
There were two very different visual systems emerging.
One was the window.
There is an ordinary domestic world — table, chairs, familiar rural surroundings.
And then there is the sky outside.
The house stays a house.
The sky stops behaving.
That became a visual boundary between the ordinary world and something I couldn’t quite name.
Then there was the walking woman.
Low shots of legs in heels moving along asphalt, crosswalks and highway markings.
Different shoes.
Same purposeful gait.
We never really get her face.
We don’t know where she is going.
She isn’t a conventional character.
She is simply moving through the world.
And then there are the tiny solitary figures appearing in the surreal landscapes — walking through fog, standing beneath impossible structures, trying to navigate places that seem to have been constructed by a dream with incomplete instructions.
I didn’t consciously design all of that symbolism.
I noticed it afterward.
That’s the part that fascinates me.
6/9
Step 4: Then somebody psychoanalyzed my video.
After the film was finished, I had the whole thing analyzed psychologically.
I expected some interesting observations.
I did not expect to recognize the movie.
The analysis pointed out that the film keeps splitting itself between the window and the stride.
The window looks outward at something increasingly uncanny.
The walking body keeps moving through ordinary space.
The solitary figures occupy the space between those two systems.
And then there is the voice:
“Wait. What was that?”
“Never mind. Keep going.”
Later:
“Wait a minute. I’ve got another question.”
And finally:
“Play it again.”
That last line suddenly made the whole structure click for me.
The film isn’t answering a question.
It is performing the act of having one.
Something appears.
I don’t understand it.
I wave it away.
I keep moving.
Then I come back and look again.
7/9
Step 5: The happy accident
There is even a darker quality in the finished imagery than I originally intended.
Some of the sky formations are beautiful.
Some look ominous.
One image can almost read as a mushroom cloud.
The final view from the domestic room can look almost picturesque until you notice what is happening on the horizon.
That was not the story I sat down to tell.
But I’m increasingly convinced that this is one of the most valuable things AI can do for me.
I can begin with an intention.
I can provide the tools with enough structure to keep me pointed in a general direction.
But the system can introduce associations I didn’t consciously choose.
Then I get to decide whether those associations are noise or whether they have something to say.
Lab note:
I don’t think AI is replacing the artist here.
It’s giving the artist something to argue with.
Sometimes I tell the machine, “No. That’s not what I meant.”
Sometimes I stop and think, “Wait a minute.”
8/9
Step 6: The film became more honest than the plan
The original concept was about curiosity and play.
The finished film still contains those things.
But it also contains uncertainty.
It contains movement without a destination.
It contains beauty sitting uncomfortably close to catastrophe.
It contains a perfectly ordinary room trying to remain ordinary while the horizon becomes impossible.
And it never resolves any of it.
That may be the most important thing I learned from making it.
I had originally thought the ending was just a playful little button:
“Wait a minute…
I’ve got another question.
…Never mind.
Play it again.”
Now I think it is the entire film compressed into four lines.
The question isn’t finished.
The movie doesn’t finish the question.
The movie is the question.
And asking to play it again is not an afterthought.
It is the natural conclusion.
9/9
The real lesson
I went into NOT IN MEMPHIS thinking I was building a visual metaphor.
What I actually built was an associative machine.
The music asks.
The images answer.
The answers create more questions.
The questions lead somewhere strange.
Then the whole thing comes back around to the room where it started.
That is probably a pretty good description of how my brain works anyway.
I start with one idea.
Three hours later I’m researching something completely unrelated.
Five hours later I’ve made a video about it.
At some point I ask myself how I got there.
Then I usually make another video.
I like that AI lets me work this way.
Not because it magically creates art for me.
It doesn’t.
I am still the one deciding what survives, what gets thrown away, what gets rebuilt and what suddenly makes me stop and say, “Wait. What was that?”
The machine supplies possibilities.
I supply judgment.
And every once in a while, the two of us stumble into something neither of us quite expected.
That’s the good stuff.
So…
if you watch NOT IN MEMPHIS and think you understand it the first time, you may be missing the point.
Play it again.
Steve Teare
video alchemist
TerminallyBored.Monster
Palouse, Washington USA