My 3-Step Workflow For Ultra-Realistic AI Ads
Establishing a structured asset-first workflow with dedicated LLM prompting tools is essential for producing cohesive, high-production commercial spots without burning through generation credits.
What this lesson covers
Thomas Lundström breaks down a three-step AI filmmaking pipeline inside Higgsfield Cinema Studio to produce ultra-realistic product commercials in a single day. By leveraging reference assets, a specialized Claude skill for shotlist prompting, and controlled video generation, creators can achieve consistent multi-shot ads while keeping credit costs minimal.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 03:15 ↗
Save generated product and character reference sheets as distinct elements in Higgsfield Cinema Studio to enable automatic tag-linking during video generation.
- 04:05 ↗
Omit the face from full-body reference sheets so AI video models maintain precise facial consistency across different shot angles and outfits.
- 04:48 ↗
Generate environment backgrounds with Cinema Studio's Cinematic Locations tool to establish set assets cheaply before rendering video clips.
- 05:36 ↗
Upload a screenshot of the Cinema Studio elements tab to Claude so the Shotlist Director skill automatically tags element names inside generated prompts.
- 06:28 ↗
Limit each 15-second video prompt to three or four shot cuts maximum to preserve visual stability and render quality.
Workflow outlined in the video
- Generate asset reference sheets for characters, props, and locations inside Higgsfield Cinema Studio.
- Crop or edit character reference full-body sheets to exclude the face, preserving separate close-up face references for consistency.
- Save each asset as a tagged element within the Cinema Studio project space.
- Install the Shotlist Director skill in Claude and upload screenshots of the project elements along with the commercial script prompt.
- Copy the automatically formatted shotlist prompts from the HTML export into Cinema Studio's video generator.
- Render initial video passes in 1080p to monitor quality and continuity before committing high-credit 4K renders.
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
11,869 views captured 24 October 2026. ReelStack recommendation score: 57/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 3.66× views vs 4 other channel videos observed at a similar age. View-evidence factor 92% (views / (views + 1,000))
- 25% freshness: 1/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 100/100; likes + 4× comments, smoothed with a 500-view neutral prior