I Mixed AI With Real World Footage Using Higgsfield Genjutsu

What this lesson covers

Jahan demonstrates how to use Higgsfield's Genjutsu tool to integrate generative AI characters and visual effects into real-world camera footage. He breaks down the differences between object swap and motion transfer modes, detailing prompt structures, scaling rules, and reference image workflows.

Key takeaways from the creator

AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.

1. 01:42 ↗

   Genjutsu uses the source video to establish motion, camera tracking, and timing, requiring prompts to specify only what changes and what stays locked.

2. 02:18 ↗

   Motion transfer preserves the source timing and motion while rebuilding the scene, whereas object swap alters specific characters or objects while preserving the surrounding environment.

3. 03:08 ↗

   Prompts should begin by locking the base clip's movements, facial expressions, and camera, followed by specifying reference-linked characters.

4. 04:32 ↗

   Defining an AI character's physical scale relative to an on-screen subject (such as one-third the size of a human) prevents scaling and placement hallucinations.

5. 05:51 ↗

   Explicitly describing camera motion or choreography in the text prompt should be avoided because the source clip already dictates motion, and re-prompting it causes generation conflicts.

Workflow outlined in the video

1. Shoot live-action reference footage capturing the intended camera motion, pacing, and human performance.
2. Create consistent character reference sheets using an image generator such as GPT images 2.5.
3. Choose between Object Swap mode for inserting elements into existing environments or Motion Transfer mode for driving full character animations.
4. Draft a structured prompt that instructs the model to lock the base clip's actor performance and camera angle.
5. Specify the reference image ID, explicit placement coordinates, and scale relative to real objects in the scene.
6. Review iterations and reroll the generation if visual interactions or effect positions fail to align.

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

13,091 views captured 23 October 2026. ReelStack recommendation score: 69/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.

Read our methodology and limitations →