Complex Action and Storyboard Guide
Learn how to plan complex AI video actions with causal storyboards, 12-panel breakdowns, reference videos, and short editable segments.
Why complex action breaks
When one shot asks a character to run, jump, turn, wave, look back, and speak, the model often blends the actions into one unstable movement. The fix is to split the sequence before adding more descriptive words.
Fix the storyboard logic first
Four storyboard questions:
1. Why does the character act?
2. Where are they going?
3. What blocks the action?
4. Why does the result make sense?These questions solve causality, not visual style. Once the reason and destination are clear, camera and prompt decisions become easier to evaluate.
Two action methods
Two main methods for complex action:
Method A: 12-panel storyboard
- Best for original action design
Method B: Reference-video replacement
- Best for quickly getting a stable movement rhythm12-panel method:
1. Lock the character and environment
2. Split the complex action into 12 continuous states
3. Give each panel one intermediate pose
4. Group the panels into 3-4 video segments
5. Keep one main action chain per segmentReference-video method:
1. Find a reference with the right movement rhythm
2. Extract its key shots and action beats
3. Replace the subject with your character, scene, and style
4. Preserve the movement structure instead of rewriting it completely
5. Repair local issues and edit the segments together- Use the 12-panel method for original action, parkour, fights, dance, and emotional chains.
- Use a reference video for fast rhythm matching, while respecting rights and publishing boundaries.
Case structure
A character rushing to meet someone at a cafe is a useful example: message received, late arrival noticed, rush outside, blocked by a red light, pause at the entrance, then enter and face the waiting person. The intermediate obstacles make the final tension believable.
Split this into three or four clips: reaction and departure, street obstacle, entrance and pause, then the final encounter. Keep each clip around 3-5 seconds and connect them in the edit.
Output template
## Story goal
## Answers to the four storyboard questions
## Character and environment references
## 12-panel or reference-video plan
## Segmented video prompts
## Editing orderAction looks convulsive -> too many main actions; keep one
Character drifts -> missing character or scene anchor
Causal break -> missing reason, destination, obstacle, or result
Action is correct but unattractive -> repair shot logic before aesthetics
Reference replacement fails -> the movement structure was changed, not just the subjectRecommended workflow
1. Define the character and environment
2. Break down shots before writing a long action sentence
3. Choose the dominant action anchor
4. Lock the sequence with a 12-panel board or reference video
5. Keep one dominant action per shot
6. Generate short segments and edit them togetherIf the real problem is character or spatial drift, repair those anchors first with the scene consistency guide and the AI video director guide.