if you were waiting on a sign, this is it (master doc with skills and prompts in description)

What this lesson covers

Joey breaks down the reverse-engineered workflow behind cinematic AI shorts, leveraging Higgsfield's open-sourced Hellgrind movie assets and his own custom Claude skills. He demonstrates how multi-panel character sheets, hard-locked prompt parameters, and voice locking ensure character and object consistency across generative video pipelines.

Key takeaways from the creator

AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.

1. 03:38 ↗

   Construct a multi-panel character sheet featuring front, rear, and headshot views so video generation models lock one face to one outfit accurately.

2. 04:34 ↗

   Describe continuous handheld camera movements and micro-expressions in prompts to eliminate static AI artifacts and stiff character poses.

3. 06:21 ↗

   Create distinct element tags for character state variations (e.g., clean vs. beaten-up) to maintain visual consistency as a narrative progresses.

4. 07:24 ↗

   Implement a hard lock at the beginning of prompt strings with explicit physical dimensions for held objects to maintain proper scale across shots.

5. 08:04 ↗

   Utilize a Story Bible skill to enforce voice locking, behavioral traits, and distinct character attitudes across generative audio and dialogue scenes.

Workflow outlined in the video

1. Set up a 4-panel character sheet in your image generator containing front, rear, and facial references.
2. Define distinct character state variants (such as worn clothing or injuries) and save them as separate element tags.
3. Prepend prompt strings with clear scale constraints (e.g., precise object sizes in centimeters) to prevent spatial warping in Seedance.
4. Input specific camera dynamics (like handheld sway) and constant character micro-movements into your prompt generator.
5. Use a Story Bible LLM skill to build consistent dialogue style, character voice attributes, and behavioral traits across the scene script.

Before you use this workflow

These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.

ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.

Explore the tools

How to interpret the numbers

11,693 views captured 27 September 2026. ReelStack recommendation score: 30/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.

Read our methodology and limitations →