# Build a repeatable production workflow

Lesson 1 of [Run AI video at business scale](https://aivideogen.fyi/academy/ai-video-at-business-scale.md), 20 minutes.

One video is a project. Ten videos a month is a system. The difference is a repeatable workflow: fixed inputs, fixed prompts, fixed checks, and a log that tells you what each video cost.

## Standardise the inputs

Every recurring video starts from the same three things: a script template, a shot list template and a style sheet. The script template fixes the structure (hook, benefit, proof, close). The shot list fixes the beats. The style sheet fixes the style sentence, aspect ratio, caption style and music loudness. When inputs are fixed, outputs become comparable and improvements become measurable.

## Build a prompt library, not a prompt habit

Keep every prompt that produced a keeper in a document, organised by shot type: product reveal, benefit, texture, transition. Each entry records the prompt, the tool, the model, the credits it cost and the date. Over a month you will have a library of shots you can re-run for the next product. This is the difference between generating video and producing it.

## Decide the generate/animate/film split once

For each recurring video type, write down which shots are generated, which are animated from real stills and which are filmed or designed. A product ad might generate 3 shots, animate 2 and film 1. A weekly presenter update generates 0 and films or avatars all of them. Write the split down; do not re-decide it every time.

## Log credits and versions

Track per project: tool and plan, prompt or shot, credits or cost, takes made, keeper result. The log answers "what did this video cost" without guesswork. It also tells you when a plan's credit pool is too small: two projects in a row hitting the cap is the signal to upgrade one tier, not to buy packs twice.

## Assign owners and reviews

A workflow without an owner decays. Name who writes scripts, who runs generations, who edits and who publishes. Set a weekly production slot and a monthly review of the cost log. When the team is small, one person can hold every role; the point is that the roles are named and the rhythm is fixed, so the work survives vacations and new hires.

## Review monthly, change quarterly

AI video tools change monthly: new models, new plans, new price pages. Review your stack monthly against your log: cost per finished video, retry rate, export quality. Change the stack when the numbers say so, not when a demo is impressive. Quarterly, re-read each vendor's terms for commercial use and watermark rules, because those clauses move too.

## Sources

- [Pika pricing](https://pika.art/pricing)
- [Runway pricing](https://runwayml.com/pricing)
- [HeyGen pricing](https://www.heygen.com/pricing)
- [Synthesia pricing](https://www.synthesia.io/pricing)


Also on the [web page](https://aivideogen.fyi/academy/ai-video-at-business-scale/build-a-repeatable-production-workflow).
