From Michelin's Bibendum to Flo from Progressive Insurance, brands have always understood the power of a recognisable character. AI avatars promise to make this scalable � but most fail for one reason: they don't look like the same person twice.
01 The consistency problem
Generate ten AI avatar videos with the same prompt, and you'll likely get ten subtly different faces. Different jawlines, inconsistent skin tones, shifting eye shapes. Audiences notice � even if they can't articulate why something feels off.
The uncanny valley isn't just about realism; it's about reliability. A character who changes appearance between videos breaks the trust that recurring characters are designed to build.
02 Building a character system
At Basis Media, we treat AI characters the way animation studios treat character sheets � as documented, locked assets with defined parameters:
- Visual bible � Reference images from multiple angles, defining face shape, hair, wardrobe, and colour palette.
- Locked model weights � Fine-tuned or LoRA-trained models that reproduce the character consistently.
- Voice profile � A consistent AI voice clone or selected voice actor with defined tone and pace.
- Personality brief � How the character speaks, what they care about, what they wouldn't say.
- Motion guidelines � Approved gestures, expressions, and movement patterns.
03 Storytelling with AI characters
The most effective AI brand characters aren't visual novelties � they're narrative devices:
- The explainer � Breaks down complex products in a friendly, approachable way.
- The guide � Walks audiences through tutorials and onboarding sequences.
- The host � Anchors a content series, podcast, or social show.
- The ambassador � Represents brand values across markets and languages.
"A character isn't a gimmick. It's a commitment to show up the same way, every time, until the audience trusts who they're looking at."
� Basis Media
04 Avoiding the uncanny valley
Technical consistency alone doesn't guarantee audience acceptance. Principles that help AI characters feel intentional:
- Stylise deliberately � Illustrated or cinematic characters often feel more trustworthy than hyper-realistic ones.
- Be transparent when appropriate � Audiences accept AI characters when brands are honest about the approach.
- Limit motion complexity � Simple, well-executed movement beats ambitious but glitchy animation.
- Pair with human context � AI characters work best alongside real footage and environments.
- Iterate with audience feedback � Test reception before scaling across campaigns.
Key takeaway
A well-built AI character becomes a production asset that appreciates over time � every video adds familiarity, and localisation becomes a voice swap rather than a full rebuild.