Opus 5.5 in Blender Is INSANE for Al Commercials
Shows AI filmmakers how to bridge LLM agentic control with 3D blocking in Blender to guide generative video models, ensuring camera angles and staging match intent before committing render credits.
What this lesson covers
Thomas Lundström explains how to direct a 30-second AI commercial inspired by viral social media trends. He details creating character assets, automating Blender previz sequences using Claude Opus 5.5 via MCP, and rendering shots using Seedance 2.5 draft mode.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 02:12 ↗
Brainstorming commercial concepts can be streamlined by dictating raw ideas directly into Claude to structure a draft narrative, scenes, and taglines.
- 03:04 ↗
Establishing product, location, and character sheets upfront and saving them as named elements in Higgsfield enables automatic prompt tagging in later steps.
- 03:32 ↗
Claude Desktop paired with the Higgsfield MCP plugin allows Opus 5.5 to directly set up and iterate on 3D blocking and camera previz in Blender.
- 04:41 ↗
Iterating 10 to 15 times on the 3D blocking previz before video generation significantly improves spatial consistency and camera framing in final shots.
- 05:33 ↗
Using Seedance 2.5's draft mode permits testing video generations at lower resolution to catch errors before upscaling identical results to 1080p.
Workflow outlined in the video
- Dictate rough commercial ideas into Claude to produce a structured narrative script, tagline, and scene breakdown.
- Generate character, product, and location reference sheets and register them as elements in Higgsfield for auto-tagging.
- Set up the Claude Desktop app with the Higgsfield MCP connector and a blocking skill linked to your Blender project folder.
- Direct Claude Opus 5.5 to build and refine the 3D camera blocking sequence in Blender until shot angles and pacing are locked.
- Export the Blender previz sequence and upload it as a reference video in Higgsfield alongside the shotlist prompt.
- Run initial generations using Seedance 2.5 draft mode to evaluate movement and prompt errors before re-rendering final takes in 1080p.
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
Put the lesson in context
When to use Blender previsualization for AI video →How to interpret the numbers
16,689 views captured 12 October 2026. ReelStack recommendation score: 67/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 1.95× current views vs the channel's recent median; same-age history is not yet available. View-evidence factor 94% (views / (views + 1,000))
- 25% freshness: 92/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 30/100; likes + 4× comments, smoothed with a 500-view neutral prior