From The Optimization Film by Dawn Foster · Entertainment

Production Plan

Published

 

Given what we already developed for The Optimization—the 12–15 minute screenplay, the 33-shot production script, and the Master Character & Style Prompt, I would not try to generate the entire movie in one AI-video prompt. The practical approach is to produce it as a sequence of controlled shots and then assemble those shots into a finished film.
The good news is that the material you already have is unusually well suited to an AI-assisted production workflow.

The production approach I'd recommend

Think of it as a miniature film production pipeline:
Screenplay → Shot list → Character/style bible → Generate shots → Voice/dialogue → Sound design → Edit → Color grade → Final film
For The Optimization, I'd break the work into these stages:

1. Lock the visual "bible"

Your Master Character & Style Prompt should become the source of truth for every generated shot.
In particular, keep these elements absolutely consistent:
  • A-17: 6'2", slim humanoid, matte titanium/dark graphite body
  • Faceless design with subtle white-blue optical eyes
  • Precise, mechanical movements that gradually become more human
  • Maya: 35, dark-haired resistance leader
  • AI remains disembodied—no evil robot/AI villain
  • The AI-controlled world is beautiful, clean, quiet and seemingly perfect
  • Photorealistic cinematic appearance
  • 2.39:1 aspect ratio
  • Evolution of color and lighting as the story progresses
The biggest challenge with current AI video isn't generating an impressive shot. It's generating shot #17 that still looks like shot #4's A-17.
So I would create a small set of master reference images first:
  • A-17 full body
  • A-17 face/eye close-up
  • A-17 side profile
  • A-17 walking
  • Maya portrait
  • Maya full body
  • Maya in the resistance cave
  • The optimized city
  • The resistance cave
  • A-17 and Maya together
Those become your visual references throughout production.

2. Generate the movie shot-by-shot

Your existing production script already gives us the 33 shots.
I'd treat every shot as an individual AI video generation task.
For example, rather than asking:

"Create The Optimization."
we'd create something like:
SHOT 07 — A-17 walks through the optimized city
Then combine:
Master Style Prompt + Character Reference + Shot-specific Prompt + Camera Direction + Motion Direction
The result might specify:
  • 6-second duration
  • slow tracking shot
  • A-17 walking through the city
  • pedestrians moving naturally around him
  • no one looking at him
  • immaculate futuristic architecture
  • morning light
  • restrained cinematic color palette
  • photorealistic
  • 2.39:1
  • no text
  • no logos
  • no dialogue
Then we'd generate several versions and select the best one.

Don't try to make shots too long.

For AI-generated video, I'd generally favor 4–8 second shots rather than trying to generate 20–30 second sequences.
A 12–14 minute film can therefore be assembled from a surprisingly large number of short clips.
Your 33-shot script may ultimately become 60–100 individual clips once we break some shots into coverage.

3. Use image-to-video rather than text-to-video whenever possible

This is probably the most important production recommendation.
For recurring characters such as A-17 and Maya:
Generate the still image first → approve it → animate it.
That gives us much better control over continuity.
For example:
A-17 reference image
↓
A-17 standing in city
↓
Animate:
A-17 slowly walks toward camera while pedestrians pass behind him.
That is much more controllable than asking an AI video model to simultaneously invent:
  • A-17
  • the city
  • his appearance
  • his movement
  • the camera
  • the lighting
from scratch.

4. Pick a video-generation platform

There are several good candidates, and the landscape changes quickly. I'd evaluate the current versions of:
I wouldn't choose based simply on which produces the prettiest demo.
For The Optimization, I'd specifically test:
character consistency + image-to-video + cinematic camera movement + control over duration + ability to maintain a reference character.
Those characteristics matter much more than raw image quality.

5. Create the voices separately

I would not rely on the video generator for dialogue.
The dialogue should be generated independently and then synchronized to the video.
There are two particularly important voices:

A-17

The voice should evolve.
Early A-17:
precise neutral almost synthetic emotionally flat
Later A-17:
still restrained but increasingly human subtle hesitation emotional weight
That evolution is one of the most important storytelling devices in the movie.

The AI

The AI voice should be:
  • calm
  • intelligent
  • reassuring
  • completely unemotional
  • never sinister
  • never "robotic villain"
That's crucial because the AI genuinely believes it has solved humanity's problems.

6. Sound design may matter more than the visuals

This is where The Optimization can become much more cinematic.
Don't make the soundtrack just:
music + dialogue.
Build a sound world.
For example:

Optimized civilization

Almost unnaturally quiet.
  • distant transportation
  • soft ambient machinery
  • subtle electronic tones
  • footsteps
  • controlled environmental sounds

Resistance world

Much more organic.
  • wind
  • dripping water
  • breathing
  • rocks
  • fire
  • distant machinery
  • imperfect human sounds
The audience should feel the difference before they consciously understand it.
And A-17's movement can have a subtle mechanical sound early in the film that gradually disappears.

7. Music should evolve with A-17

I'd use a restrained cinematic score rather than a conventional Hollywood action soundtrack.
The music could progress roughly like this:
Act I — Perfection
Minimal electronic/ambient music.
Act II — Observation
More melodic elements begin appearing.
Act III — Humanity
Organic instruments gradually enter.
Ending
The music should become emotionally ambiguous rather than triumphant.
The central question isn't:
"Did humanity defeat the AI?"
It's:
"Is an imperfect life actually worth more than a perfectly optimized one?"
The music should reinforce that.

8. Edit everything in a conventional video editor

Once the AI clips are generated, the movie becomes a conventional filmmaking/editing project.
I'd use something like:
For this project, DaVinci Resolve would be particularly attractive because you can handle:
  • editing
  • dialogue
  • sound effects
  • music
  • color grading
  • titles
  • final mastering
in one application.

9. Build the movie in passes

Don't try to finish one scene completely before moving to the next.
I'd use a professional-style workflow.

Pass 1 — Storyboard

Generate still frames for all 33 shots.
Create a rough animatic using those stills.
At this point you can watch the entire movie.
You'll discover problems such as:
"This scene is too long."
or:
"We don't need this shot."
or:
"The transition from the city to the cave isn't strong enough."
Fix those before generating hundreds of video clips.

Pass 2 — Video

Once the animatic works, generate the actual moving shots.
At this stage we're solving:
  • movement
  • camera
  • continuity
  • performance
  • transitions

Pass 3 — Dialogue

Add the final A-17, Maya and AI voices.

Pass 4 — Sound

Add:
  • ambience
  • footsteps
  • machinery
  • cave sounds
  • environmental effects
  • subtle robot sounds
  • transitions

Pass 5 — Music

Add the score underneath everything.

Pass 6 — Color

This is especially important for your film.
The visual progression can become part of the storytelling.
Beginning:
Cool, clean, controlled.
↓
Middle:
Subtle warmth begins entering the imagery.
↓
Resistance:
Earthier, imperfect, organic.
↓
Ending:
A deliberate mixture of the two worlds.

10. The actual production workflow I'd use for you

Because you already have the screenplay and production script, I would turn them into a production spreadsheet.
Something like:
Shot
Scene
Characters
Location
Duration
Camera
Image Prompt
Video Prompt
Dialogue
SFX
Music
Status
01
1
—
City
6s
Aerial
✓
✓
—
✓
✓

02
1
A-17
City
7s
Tracking
✓
✓
—
✓
✓

03
1
A-17
City
5s
Close
✓
✓
A-17
✓
✓

...











Then each shot becomes a little production package.

Each shot gets:

1. Reference image
2. Image-generation prompt
3. Video-generation prompt
4. Negative/avoid instructions
5. Camera instructions
6. Character performance instructions
7. Dialogue
8. Sound effects
9. Music cue
10. Continuity notes
That makes the project manageable.

An important change I'd make to our existing production script

I'd take the existing 33-shot production script and turn it into a much more detailed AI production bible.
For each shot I'd add:

VISUAL

What the audience sees.

CAMERA

Lens/look, camera position, movement, framing.

CHARACTER

Exact A-17/Maya appearance and emotional state.

ACTION

Exactly what happens during the shot.

ENVIRONMENT

Architecture, weather, lighting, background activity.

MOTION

What is moving and how quickly.

DIALOGUE

Exact spoken words.

SOUND

Dialogue + environmental sound + SFX.

MUSIC

Music cue and emotional purpose.

TRANSITION

How the shot connects to the preceding/following shot.

AI PROMPT

The actual prompt to paste into the generation tool.
That document effectively becomes your AI director's shooting script.

The biggest challenge: A-17

A-17 is actually the hardest part of the entire project.
Not because the character is complicated, but because he appears throughout the movie.
If his:
  • proportions
  • head
  • eyes
  • shoulders
  • armor
  • material
  • color
  • height
change from shot to shot, the audience will immediately notice.
I'd therefore spend disproportionate effort at the beginning creating a character reference package for A-17.
I'd establish:
A-17 Master Reference
and then create:
  • front
  • ¾ front
  • side
  • rear
  • close-up
  • full body
  • walking
  • sitting
  • interacting with Maya
  • emotional-state references
Once those are locked, every subsequent generation references them.

You can actually make this yourself

You don't need a Hollywood production company.
The project can realistically be a one-person AI filmmaking project, with you acting as:
Director + Producer + Writer + Editor
and the AI tools providing:
Cinematography + Actors + VFX + Production Design + Voice Actors + Music + Sound Elements
The important thing is to treat the AI as a production crew rather than as a magic "make my movie" button.

What I would do next

Since we already have the screenplay, 33-shot production script, and Master Character & Style Prompt, I think the most productive next step is to create a complete AI Production Plan for The Optimization.
That would take the existing 33 shots and turn them into something you can actually execute:
33 shots → individual AI prompts → reference images → video prompts → dialogue → voice instructions → SFX → music cues → editing instructions → recommended AI tool for each task.
We could start with Shots 1–5 as a production prototype, generate the reference images and prompts, and establish the exact workflow before doing the remaining 28 shots. That would substantially reduce the risk of spending money generating a movie that later has continuity problems