Seven seconds.
That was all it took.
Seven seconds of a woman saying something she probably shouldn’t have said out loud.
“I think I like you more than I meant to.”
Then nothing.
Not rejection. Not acceptance.
Just that little blue mark on a phone saying: Seen.
That was the seed for Seven Seconds, a ten-minute AI-assisted short about vulnerability, waiting, and the astonishing amount of emotional damage we can inflict on ourselves while staring at a phone.
And, as usual, I started with a simple idea and somehow ended up building an entire tiny film studio in my computer.
Because apparently I have hobbies.
2/9
The story
The basic story is deceptively simple.
Mara has been seeing Daniel for several weeks. He’s kind. Reliable. Thoughtful. He remembers things she tells him. He shows up.
In other words, he has done absolutely nothing to deserve her anxiety.
Unfortunately, anxiety doesn’t require evidence.
After a good evening together, Mara sends Daniel a seven-second voice message telling him she likes him more than she intended.
Daniel listens.
And then his day gets away from him.
Mara spends the next several hours slowly dismantling herself.
She almost sends an apology.
She deletes it.
She checks her phone.
Checks it again.
Checks it again, because apparently the human brain becomes a surveillance drone when someone doesn’t text back.
Eventually they meet for dinner.
And instead of allowing the misunderstanding to grow, Mara does something much harder.
She tells him what happened inside her.
That becomes the real story.
Not “Will he like her?”
But:
Can she tell the truth about what she feels without immediately trying to take it back?
3/9
The first problem: making people
This is where things got interesting.
I wanted Mara and Daniel to feel like actual human beings rather than the usual suspiciously perfect AI romance models.
No porcelain skin.
No impossibly symmetrical faces.
No “I have never experienced a bad hair day in my entire life” nonsense.
I created individual character references for both of them in Midjourney.
Mara became the visual anchor of the film: dark brown wavy shoulder-length hair, hazel eyes, natural skin texture, understated clothing, intelligent face, and just enough vulnerability around the eyes to make tiny emotional changes visible.
Daniel got his own reference: late thirties, short dark brown hair, light stubble, lean build, warm eyes, understated clothes, approachable rather than Hollywood handsome.
Then I discovered something important.
Midjourney is not particularly interested in respecting my desire for two emotionally complicated human beings to remain two emotionally complicated human beings.
Put both character references into a shot and things can get… creative.
A beard may migrate.
A blouse may appear where no blouse should be.
Suddenly Daniel has discovered eyeliner.
Mara has apparently grown a five-o’clock shadow.
AI is many things.
A continuity supervisor it is not.
4/9
The one-character rule
So I changed the strategy.
One identifiable character per frame.
One character reference per prompt.
That became one of the most important decisions in the entire production.
When Mara and Daniel are together, I don’t need both faces staring into the camera.
In fact, I don’t want that.
Instead, I shoot them like a real film.
Mara looks toward Daniel.
Cut.
Daniel looks toward Mara.
Cut.
Mara reacts.
Cut.
Daniel answers.
The audience does the rest.
When both characters appear in the same frame, one becomes secondary — a shoulder, a silhouette, a hand, the back of a head.
And whenever both are present, Mara gets priority as the recognizable character.
Lab note:
This turned out to be more than a technical workaround.
It made the movie better.
The limitation forced me toward actual cinematic language.
Shot/reverse-shot.
Eyelines.
Reaction shots.
Hands.
Silhouettes.
Negative space.
All the things filmmakers have been doing forever because you don’t actually need two perfectly visible faces in every frame to understand two people are having a conversation.
5/9
Eighty-five little pieces of a human being.
The finished film needed approximately 85 still images.
That sounds like a lot.
It is.
Especially when every image has to remember what came before it.
Mara can’t suddenly have a different hairstyle at dinner.
Daniel can’t age twelve years between two lines of dialogue.
The restaurant can’t change from a candlelit romantic room into an airport lounge.
So I built the shot list first.
Then I converted every shot into a Midjourney prompt.
Each prompt specified:
• camera angle
• lens
• composition
• character action
• emotional state
• environment
• lighting
• color palette
• cinematic style
• character reference
The lighting became especially important.
Morning scenes use soft natural window light.
The restaurant lives in warm practical light and candlelight.
The station scenes shift into cool evening ambient light with warm yellow fixtures.
The visual world changes as the emotional world changes — but never arbitrarily.
Lab note:
AI image generation is surprisingly good at creating beautiful pictures.
It is considerably less good at remembering that beautiful picture number 47 is supposed to live five minutes after beautiful picture number 46.
6/9
The music had to know when to shut up.
The narration carries most of the emotional weight in “Seven Seconds.”
That meant I didn’t want a giant Hollywood score trying to tell the audience what to feel.
So I designed three roughly three-minute instrumental pieces.
The first is “The Space Between.”
Minimal piano. Soft synth textures. A little pulse. Lots of air.
It belongs to the waiting.
The second is “Say It.”
Warmer piano and Rhodes, subtle percussion, gentle bass and airy textures.
It gradually moves from uncertainty toward understanding.
The third is “Start From the Beginning.”
Warmer. More rhythmic. More open.
The musical idea resolves.
The trick was to give all three pieces the same little musical DNA.
The first version of the motif feels unresolved.
The second begins to open it.
The third finally lets it land.
Nothing shouting, “THIS IS THE HAPPY PART.”
Just a subtle feeling that the emotional ground has shifted.
That is the kind of thing I like AI music for.
Not replacing a composer.
Giving me another instrument to play.
7/9
The real experiment was emotional.
The technology is fun.
Midjourney is fun.
Voice synthesis is fun.
Editing 85 still images into a ten-minute moving story is definitely a peculiar kind of fun.
But none of that is actually the interesting part.
The interesting part is Mara.
Because Mara isn’t really afraid that Daniel doesn’t like her.
She’s afraid that she has exposed herself and given someone else the power to hurt her.
That’s a different fear.
And it’s one I suspect a lot of us recognize.
The little apology she almost sends — “Sorry. That was probably a little much” — is the entire psychological problem in one sentence.
She isn’t apologizing because she did something wrong.
She’s trying to abandon her own feelings before someone else gets the opportunity to reject them.
That hit me.
Because the opposite of vulnerability isn’t strength.
Sometimes it’s preemptive disappearance.
8/9
The happy accident
The story started as a fairly ordinary romantic misunderstanding.
But the more I worked on it, the more I became interested in something underneath the romance.
Mara doesn’t become secure because Daniel gives her the perfect response.
He doesn’t magically cure her anxiety.
He simply explains what happened.
He says, essentially:
I heard you.
I cared.
I was busy.
I should have told you.
And that is enough.
Later, when Mara sends another voice message about the new project she’s been asked to lead, Daniel responds quickly:
“I listened. I’m walking into a meeting, but I want to hear everything tonight.”
That tiny sentence is the real ending.
Mara doesn’t spiral.
She puts the phone facedown.
And gets back to work.
That is growth.
Not becoming fearless.
Learning that you don’t have to abandon yourself while you’re waiting.
9/9
The real lesson
I started “Seven Seconds” thinking I was making a little story about a woman waiting for a text.
I ended up making a story about presence.
About what happens when we stop trying to control another person’s response and start paying attention to our own response.
AI was useful here because it forced me to work differently.
I couldn’t simply photograph two actors and tell them what to do.
I had to construct the emotional continuity piece by piece.
A face.
A hand.
A phone.
A glance.
A pause.
A seven-second voice note.
Eighty-five images later, the pieces began to behave like a person.
That’s the strange thing about this kind of filmmaking.
The AI doesn’t understand Mara.
It doesn’t understand anxiety.
It doesn’t understand love.
I do the understanding.
The machine provides the pixels.
I’m just the guy in the middle trying to convince all of them to tell the same story.
That’s the job of the video alchemist.
And sometimes, if we’re lucky, the pixels start feeling human.
TL;DR:
Sometimes the most frightening thing you can say is “I like you.”
Sometimes the hardest thing to learn is how to wait for the answer.
And sometimes seven seconds is enough to make an entire story.
Steve Teare
video alchemist
TerminallyBored.Monster
Palouse, Washington USA