# Zero Page AI Studio

> An automated video production machine that mixes real footage and AI: concepts, shot lists, and platform-native AI video prompts, all grounded in your real rooms and footage — and it learns from what performs.

Zero Page AI Studio is an automated video production pipeline for filmmakers:
real footage and AI-generated shots are co-inputs to the same concept,
and the system learns from what performs.

## What "grounded" means here

Generative video tools invent the world: the rooms, the light, the whole
scene come out of the model. This starts from reality instead. You
photograph the spaces you actually have; vision models describe each
one's geometry, light sources, textures, and constraints; and every
concept, shot list, and AI prompt is generated from that record — so AI
shots extend real rooms and cut against real footage without a seam.
Grounding shapes the generation; mismatches surface as visible
advisories for the filmmaker, not silent rejections.

## The pipeline

1. **Rooms** — photograph a space; a vision model records what it can and
   cannot shoot.
2. **Ideas** — a slate of concepts generated from those rooms in one
   call, so they vary against each other.
3. **The pick** — a human keeps a few. That choice is recorded against
   the prompt hash that produced it, so prompt changes are measured
   rather than argued about.
4. **Shot list** — each kept idea becomes a plan mixing camera shots
   (shot solo: tripod, self-timer, POV mount) and AI shots, each AI shot
   carrying a paste-ready, platform-native prompt.
5. **Cut list** — after the shoot, footage is ingested and an edit spec
   is produced: real clips by filename, generated clips by description,
   in one cut plan executed by hand in the edit bay.
6. **The loop** — posted videos and their analytics feed back in,
   compared at equal age, and proven winners ground the next slate.

## Capabilities

- Concepts mixing any number of real and AI shots.
- AI video prompts for Runway, Veo, Kling, Seedance, LTX, and Wan,
  anchored by a Midjourney reference still — one controlled shot
  vocabulary, so prompts stay reproducible.
- Retrieval-grounded ideation over a PostgreSQL + pgvector reference
  library, including the filmmaker's own proven winners, so generated
  work inherits a house style and a performance record rather than a
  model's default taste.
- Every human choice recorded against its prompt hash — the measurement
  that makes self-improvement real rather than claimed.
- Advisory checks on cut plans (in/out points inside real clip
  durations, runtime in range) — visible signal, never a gate.

## Compared to generative video tools

| | Generative video (Runway, Flow, Sora) | Zero Page AI Studio |
|---|---|---|
| Source of truth | The model | Your photographed rooms and shot footage |
| Generated shots | Every shot | As many as the story needs, cut against real footage |
| Feedback | None | Posted-video analytics ground the next slate |
| Output | Clips | Shot lists, platform-native prompts, and a cut plan you execute |

## Pages

- [Zero Page AI Studio](https://zeropage-studio.fly.dev/): An automated video production machine that mixes real footage and AI: concepts, shot lists, and platform-native AI video prompts, all grounded in your real rooms and footage — and it learns from what performs.

## Attribution

Built by Michael Massaad, an actor, editor, and colorist shooting
Blackmagic 6K, as the pre-production tool for Zero Page Films. The
aesthetic constraints in the system are a working filmmaker's, not a
model's defaults.