How to Prompt Seedance 2.5 for Al Filmmaking (Full Workflow)
Rigorous pre-production script chatting with LLMs drastically reduces costly video generation trial-and-error in Seedance 2.5.
What this lesson covers
Thomas Lundström outlines his end-to-end AI commercial workflow using Claude for concept drafting and Seedance 2.5 for multi-shot video generation. He details how to structure character reference sheets and design pattern-interrupt commercial concepts that maintain audience engagement.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 02:43 ↗
Structuring video ads with a pattern interrupt—transitioning from high-speed glossy visuals to a grounded comedic beat—enhances viewer retention.
- 03:38 ↗
Restricting 30-second Seedance 2.5 prompts to 12-15 cuts maximum gives the model sufficient temporal room to focus on individual shot quality.
- 04:52 ↗
Spending time refining script details and taglines in text-based LLMs prevents credit burn caused by rerunning video prompts.
- 06:01 ↗
Keeping character reference sheets simple—a tight facial close-up and full-body front/back views—yields cleaner generation results than overcrowded sheets.
- 06:48 ↗
Removing the head from full-body front reference images eliminates facial conflict for Seedance 2.5 during character video generation.
Workflow outlined in the video
- Brainstorm commercial narrative arcs and taglines with Claude before generating any visual assets.
- Limit 30-second multi-cut prompt specifications to no more than 12-15 distinct shot cuts.
- Generate facial and body reference sheets in Cinema Studio using the Soul Cinema model.
- Edit out the face from full-body front reference sheets to prevent facial reference confusion.
- Assemble environment reference images in Cinema Studio to lock in lighting and color grading before initiating video generations.
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
28,340 views captured 27 September 2026. ReelStack recommendation score: 63/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 5.42× current views vs the channel's recent median; same-age history is not yet available. View-evidence factor 97% (views / (views + 1,000))
- 25% freshness: 11/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 74/100; likes + 4× comments, smoothed with a 500-view neutral prior