Giving users a voice: Avatar generators for presenting user verbatim and insights

Experiment

Nothing will ever beat talking to human beings. Not even AI. But, lately Iโ€™ve been exploring new ways to bring user voices into the roomโ€”especially when the room is full of stakeholders. And not just quotes on a slide. I mean really bringing users to life.

User Needs
Voice of the User
Alignment

What I tried…

Give real users a synthetic voice using their actual verbatim
I used direct user verbatim to generate synthetic HeyGen and Hedra avatars. Each quote was associated with a persona (e.g., โ€œPower Buyer with Trade-inโ€ or โ€œExperienced Dealer Agentโ€).

Video caption.

The upside: why Iโ€™ll probably keep using it

Emotional resonance
Synthetic faces + voices transform cold research artifacts into something closer to a documentary. Itโ€™s harder to ignore a frustrated โ€œuserโ€ who looks you in the eye.
Speed and control
Unlike real video editing (which is time-consuming, redacted, and often restricted by NDA), HeyGen let me compose polished, on-brand user expressions quickly.
Accessibility and consistency
Everyone hears the same tone, pacing, and clarity. That helps stakeholders focus on what is said, not how a participant stumbled through saying it.

The pitfalls: why it still gives me pause

The uncanny valley is real
Synthetic faces + voices transform cold research artifacts into something closer to a documentary. Itโ€™s harder to ignore a frustrated โ€œuserโ€ who looks you in the eye.
Risks of oversimplification
A 15-second avatar clip can flatten a complex insight into a sound bite. And if youโ€™re not careful, it starts to feel like youโ€™re scripting users, not representing them.
Ethical ambiguity
Even with user consent and paraphrased language, thereโ€™s a weirdness to creating a โ€œfaceโ€ for someone who never appeared on camera. Itโ€™s respectfulโ€ฆ but also performative. Itโ€™s a line Iโ€™m still defining.

User verbatim communicated using HeyGen

User verbatim communicated using Hedra

Archetype: Vehicle Buyer (with Trade-in)

Persona: Early Adopter / Frequent Trader
Name: Nick Patterson

Actual user verbatim communicated using HeyGen

Demographics

Age:

62

Occupation:

Personal Wealth Mgmt

Income:

$195K annually

Location:

Savannah, Georgia

Education:

Masters in Finance

Status:

Married

Vehicle Trading Behavior

Frequency:

Every 12-15 months

Motivation:

Enjoys latest tech, vanity

History:

3 vehicles in 3 years

Spends:

$85K – $125K

Prefers:

New and Pre-owned

Primary Frustrations

Trade-in/Purchase Coupling:
Forced to navigate complex dual flows simultaneously

Inaccurate Valuations:
Online estimates don’t match reality (part of 40% who research online, 10% who get actual offers gap)

Information Overload:
40-56 clicks through Carvana-style flows while trying to buy AND trade

Paperwork Burden:
Spends 2+ hours at dealerships on documentation

Memory Gaps:
Can’t recall specific trim details, options, maintenance history for trade-in

Primary Goals

Separation of Concerns:
Trade current vehicle on its merits, buy next vehicle on its merits

Accurate Trade Valuations:
Real-time, data-driven assessments based on actual vehicle condition

Purchase Efficiency:
Complete more transaction steps online, minimize dealership time

Information Transparency:
Access to complete vehicle history and predictive insights

Digital & Technology Profile

Early Adopter:
First to try new automotive apps and connected car features

Device Usage:
Heavy multi-device user (smartphone primary, desktop for research, tablet for comparisons)

Connected Vehicle Enthusiast:
Actively uses vehicle apps, OTA updates, remote features

AI Comfort Level:
High – regularly uses AI tools in work and personal life

Data Sharing Attitude:
Privacy-conscious but pragmatic – willing to share vehicle data for tangible value

Technology Adoption

Vehicle Data Collection:
Willing to share driving patterns, maintenance data, location for accurate valuations

AI-Driven Insights:
Wants predictive analytics on vehicle value, optimal trade timing

Automated Processes:
Interested in “smart car selling itself” concept

IoT Integration:
Uses connected features, comfortable with vehicle data APIs