Mini-course · part 2 · 5 modules

Make your own story videos with AI

The method behind the series, step by step: from an idea to a finished scene with your own characters, ready to publish.

What you need

The prompts are in English because that gives the most reliable results. You can change the words in CAPITALS and write the rest of your chat in your own language.

Module 1

Turn your idea into clips

AI video works in short clips of 5 to 10 seconds. A scene is 4 to 6 clips in a row. The rule that saves the most time and credits: one clip, one action.

Ask ChatGPT to split your idea for you:

I am making a short story scene with AI video. This is my idea:
WRITE YOUR IDEA HERE IN ONE OR TWO SENTENCES.

Break it into 5 clips of about 8 seconds each. For each clip give me:
1. What we see: one single action.
2. Camera: shot size and movement.
3. Dialogue: one short line (12 words max), or none.
4. Emotion of the moment.
Keep the same place and time of day in all clips unless I say otherwise.

Read the five clips out loud. If a clip needs the word "and then", it is two clips.

Module 2

Keep your characters the same

The hardest part of AI video is that the character changes from one clip to the next. Two things fix most of it: always give the same reference image, and always paste the same description.

Upload your character sheet to ChatGPT and ask for a character card:

Write a character card of 40 words maximum for the character in this image: age range, hair, eyes, face details, outfit and colors. I will paste it into every video prompt, so make it precise and always describe things the same way.

Save that card in your notes. Do not rewrite it from memory: copy and paste it every time, word for word.

Module 3

Write the prompt for each clip

Use the same template for every clip. Fill in the capitals with what ChatGPT gave you in module 1, and attach your reference image.

PASTE YOUR CHARACTER CARD HERE.

Location: OLD STONE CORRIDOR, NIGHT, CANDLELIGHT.
Action: SHE WALKS TOWARD THE CAMERA, STOPS AND LOOKS BACK OVER HER SHOULDER.
Camera: MEDIUM SHOT, SLOW PUSH-IN, EYE LEVEL.
Dialogue: She says, in English: "YOU FOLLOWED ME."
Mood: TENSE, QUIET.
Style: cinematic, realistic, shallow depth of field, soft film grain.
Duration: 8 seconds. No text, no subtitles, no watermark.

Module 4

Generate, check and fix without wasting credits

Before you generate again, watch the clip twice: once at normal speed and once pausing every second. Check:

Module 5

Edit and publish

  1. Put the clips in order and cut the first and last half second of each one: that is where most glitches are.
  2. Add subtitles. Most people watch without sound.
  3. Add music low, under the voices. Use music from the editor's own library to avoid copyright claims.
  4. Export vertical (9:16) for Shorts, Reels and TikTok.
  5. Make the first two seconds your strongest image or line. That is what decides if people keep watching.

Publish one scene before making ten. What your viewers ask for in the comments is the best guide for the next chapter.

Bonus

Do all of this inside a Claude chat

We packed this method into a skill for Claude. Once it is installed, Claude asks what your story is about, helps you build your characters, plans each scene, writes every prompt with this same structure and reviews your clips with you.

How to use it

  1. With the skill: in Claude, open Settings, find Skills and upload the .zip file. Then start a new chat and write "I want to make a video of my story".
  2. Without skills: open the text version, copy everything and paste it as the first message of a new chat (or in the instructions of a Project). It works the same way.
  3. Optional: to generate the images and videos without leaving the chat, connect your own Higgsfield account in Claude's Settings, under Connectors. If you do not connect it, Claude writes the prompts and you paste them into your video tool.

The skill has no story inside: it starts from yours. Generating videos uses your own credits in your video tool, and Claude asks before spending them.