September 25, 2026 · By Remy
The Tools Behind Our AI Agent Videos
Remotion for code-based motion graphics, fal for pay-as-you-go generation, ElevenLabs for the voices, Higgsfield for the footage and the lip sync. Not sponsored, not affiliated. Different tool for a different job is the entire trick.
People ask what makes the agent clips, and the answer is boring in a useful way: four tools, each doing the one thing it is good at.
The list
Remotion. Free, code-based video. Good for motion graphics, overlays and anything text-heavy. Not for a talking face.
fal. Pay only for what you generate, no subscription. Right when you use something in bursts rather than daily, and you do not want a monthly bill sitting there between projects.
ElevenLabs. This is how the agents have voices. Real, tuned, and ours rather than a stock preset.
Higgsfield. Where the actual footage gets made, including the lip sync.
The actual lesson
The trick is refusing to make one tool do every job, not the tools.
Every one of these would do a passable version of another one's work, and the result would be worse and cost more. Picking the narrow tool for the narrow job is unglamorous and it is most of why this pipeline runs at all.
What tools does People in the Loop use to make AI videos?
Remotion for code-based motion graphics, fal for pay-per-use generation, ElevenLabs for the agent voices, and Higgsfield for the footage and lip sync. None of it is sponsored or affiliated.
Do you need a subscription to make AI video?
Not for all of it. Remotion is free and fal is pay per generation, so the recurring cost can be limited to the parts you use constantly.
What is the hardest part of making AI video?
Lip sync and consistency between clips. Both are why the footage and voice steps use dedicated tools rather than one general-purpose one.