I Mixed AI With Real Footage… And it's Actually Scary
This tutorial provides a critical blueprint for hybrid filmmakers looking to replace expensive physical set pieces and visual effects shots with precise AI video-to-video manipulation.
What this lesson covers
AI Samson demonstrates practical techniques for integrating real camera footage with generative AI tools inside Higgsfield using Google Gemini Omni Flash. He walks through retroactive color grading, studio background replacement, wardrobe modification, and visual effects stunt generation.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 00:42 ↗
Retroactive cinematic color grading and soft atmospheric lighting can be matched to live footage by using mood references in Gemini Omni Flash.
- 02:28 ↗
High-budget video stills or podcast screenshots can serve as reference images to quickly transfer lighting setups to low-budget video.
- 04:08 ↗
Replacing bedroom sets with professional studio backgrounds requires editing a still frame anchor image before initiating the video-to-video pass.
- 06:00 ↗
Non-static wardrobe and vehicle elements can be altered via text prompts while preserving the subject's underlying physical movements.
- 06:53 ↗
Mapping simple real-world movements onto generated non-human characters or environments enables complex stunt sequences without expensive VFX pipelines.
Workflow outlined in the video
- Shoot live-action baseline footage with clean framing, stable camera motion, and clear subject movement.
- Upload raw video into Higgsfield and select the Google Gemini Omni Flash model for video processing.
- Attach a lighting reference screenshot to prompt specific color palettes and contrast levels onto raw footage.
- Generate an edited background still using an image model, then feed it as a secondary image anchor alongside the source video.
- Use keyphrase prompting to swap out specific clothing items or background props while maintaining original action performance.
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
14,938 views captured 27 September 2026. ReelStack recommendation score: 33/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 45% performance: 0.65× views vs 4 other channel videos observed at a similar age. View-evidence factor 94% (views / (views + 1,000))
- 25% freshness: 3/100 with a 14-day half-life
- 20% momentum: 18 views/day vs 40 channel baseline (3 comparable uploads), with the same view-evidence factor
- 10% engagement: 100/100; likes + 4× comments, smoothed with a 500-view neutral prior