OPINION

Human to Human, Not Avatars

Why a lab that could build avatars carries people instead.

FAMILIAR LABS RESEARCHALL ARTICLES

ABSTRACT

Agents will soon talk to people everywhere, and many will wear a face. The moment we know that face is AI, the caring part of empathy drops. So our team, researchers who could have built avatars, carries the performance a person gave into another language instead.

What the label takes away

STUDIES

9

PEOPLE

Over 6,000

WHAT DROPPED

The caring part of empathy

WHAT HELD

Feeling understood

Identical replies were rated less empathetic, supportive and authentic once labelled AI (Rubin, Perry, Goldenberg, Ong and colleagues, Nature Human Behaviour, 2025). The label is coming either way: the EU's AI Act requires realistic AI-made video and audio to be disclosed from August 2, 2026. We think the reaction is intrinsic, and actors and singers mostly stay human.

Photoreal, and empty

THE SCREEN
WHERE AI LANDS WELL
Animation: a story drawn into motion.
WHERE IT FEELS EMPTY
A photoreal actor standing in for the one who would have acted.
THE ASSISTANT
WHERE AI LANDS WELL
A cartoon guide, plainly a machine.
WHERE IT FEELS EMPTY
A photoreal face standing in for a person.
THE AD
WHERE AI LANDS WELL
A jingle under the product.
WHERE IT FEELS EMPTY
A made-up celebrity, when the point is that the person really uses it.

Art is effort and creativity; AI can be art in animation's lane, novels and manga brought to the screen with no real actor in them. Photorealism that replaces a person loses: the audience came for the human.

The music test

We want humans to sing and to rap.

A song is someone's lived experience, performed. A fully generated one is fine where nobody listens closely: an ad, a store, a background track. Video works the same way.

Gen Z follows creators for their real personalities and wants real people: streams, sport, the news, a host showing what they really use.

The lab that could build avatars

Our team researches human video generation. We carry a person's performance into another language, the outbursts and the whispers intact: 13 languages today, live broadcasts in the works.

++

We never generate a person. We never replace an actor.

More in Human to human, not AI to AI and why we exist.

QUESTIONS

Does Familiar Labs make avatars?

No. Familiar Labs never generates a person and never replaces an actor. It carries a person's performance into another language.

Why does knowing it is AI matter?

In nine studies with over 6,000 people, identical replies were rated less empathetic once labelled AI, the drop in the caring part of empathy (Nature Human Behaviour, 2025).

Where does AI belong in video, then?

In animation with no real actor in it, a cartoon guide on an app, a background track. Where a person acted, the person stays.

REFERENCES

  1. [1]Rubin, Perry, Goldenberg, Ong et al., Comparing the value of perceived human versus AI-generated empathy, Nature Human Behaviour, 2025 (Hebrew University summary)
  2. [2]Regulation (EU) 2024/1689, the AI Act, Article 50 (transparency), applies from August 2, 2026

Multimodal Translation

Familiar Labs builds generationally better dubbing with a Multimodal Localization Foundation Model (MLFM), the first foundation model for multimodal translation: one model decides what to localize across text, audio and visual, and shows you the choices. Your voice and lips, carried into new languages. They're a team of 7 with PhDs, backed by Samsung, RedBird, Gold House. Our mission

BACKED BY

  • Wheelhouse
  • Maximum Effort
  • Stanford
  • Yonsei
  • University of Waterloo