[Demo] Deploying AI doppelgangers to de-identify user research recordings
Summary
Under biometric privacy laws like BIPA and CCPA, user research recordings containing users’ faces or voices can put your company at risk for lawsuits and fines. Legal departments are increasingly requiring more stringent redaction, and in some cases banning recording outright. This comes at a high cost for UX teams who are already being asked to do more with less, as losing access to recordings can increase duplicative research effort and reduce the accuracy of results. AI offers new solutions for UX teams who want to keep research recordings longer without violating biometric privacy laws. In this demo, we’ll show how we used off-the-shelf tools to intelligently redact users’ voices, faces, and bodies in research videos. By removing biometric identifiers, you can compliantly archive research recordings indefinitely, enabling your team to mine them for insights for years to come.
Key Insights
-
•
Biometric privacy laws increasingly restrict recording and retaining user faces and voices in UX research, risking large fines.
-
•
Replacing users with AI-generated avatars preserves the richness of user recordings while protecting privacy.
-
•
Wonder Studio can segment and animate a user's body and face with a 3D avatar fully automatically, simplifying production.
-
•
11 Labs technology enables seamless voice replacement using AI-generated speech in user research videos.
-
•
Replacing the entire body, not just the face, mitigates risks from distinctive identifying marks like tattoos and future privacy laws.
-
•
Traditional redaction methods like blurring or keeping only transcripts are either insufficient or lose valuable context.
-
•
Current AI tools have usage limits, require legal reviews, and sometimes restrictive licensing affecting enterprise use.
-
•
Hosting AI models in-house offers more data control but adds technical complexity compared to cloud-based solutions.
-
•
Using synthetic avatars raises complex ethical questions about authenticity, identity representation, and users' rights to be forgotten.
-
•
Legal risk tolerance varies, so negotiating with legal teams is essential to implement AI avatar redaction in research workflows.
Notable Quotes
"User recordings are your most valuable asset but have become riskier due to biometric privacy laws."
"Even if you’re not doing facial recognition, storing face and voice data is under growing legal scrutiny."
"Blurring user faces loses context and doesn’t address voice privacy, making it an inadequate solution."
"Wonder Studio automatically segments actors, maps their movements to a 3D model, and renders a synthetic avatar video."
"Replacing the entire body with an avatar future-proofs against unanticipated identifiers like tattoos or moles."
"DeepFakes can look too realistic and might introduce new privacy risks if donor faces come from real people."
"Processing limits and licensing terms currently restrict how much video and audio these AI tools can handle."
"Ultimately, AI avatar tools offer UX teams options to keep recordings while meeting legal compliance."
"If you can’t see a human in a video, how do you know the entire conversation wasn’t fabricated?"
"Synthetic duplicates living on after data deletion raise ethical questions about users’ right to be forgotten."
Or choose a question:
More Videos
"Every piece of assistive technology is highly customized for the individual, so labs often fail to replicate users’ ideal setups."
Sam ProulxSUS: A System Unusable for Twenty Percent of the Population
September 29, 2021
"We can’t design for every possible outcome anymore. We have to anticipate fuzzy ranges of results and channel behavior accordingly."
Josh Clark Veronika KindredSentient Design, AI, and the Radically Adaptive Experience (1st of 3 seminars)
January 15, 2025
"Impact of research is not always immediate or linear; it can be a long-term, complex process to measure."
Taylor KlassmanShaping the Next Era of UX Research: Collaborative Forum
March 11, 2025
"When I gave the model my list of interview questions, it actually made for a really bizarre participant experience because the model just went down the line of the different questions."
Tara TresselInvestigating qualitative depth of AI-moderated interviews
March 10, 2026
"Communications authored for one cultural group may have limited engagement from others, underscoring the need for tailored messaging."
Angy Peterson Bob AinsburyMore Than Technology: Personalized Public Sector Experiences
December 10, 2021
"Junk in, junk out. Bad data is worse than no data at all."
Jemma Ahmed Steve Carrod Chris Geison Dr. Shadi Janansefat Christopher NashDemocratization: Working with it, not against it [Advancing Research Community Workshop Series]
July 24, 2024
"You have to hold people’s hands through your organization—not just hype about AI."
Erika FlowersAI-Readiness: Preparing NASA for a Data-Driven, Agile Future
June 10, 2025
"We’re building a program here that lasts through 2025 and 2030, so we can take risks on entry level folks."
Marc Rettig Julie Baher Phil Gilbert Nathan ShedroffDiscussion
May 14, 2015
"Learning is fractal—self-similar at different scales from individuals to organizations."
Jen BriselliLearning Is The Engine: Designing & Adapting in a World We Can’t Predict
April 16, 2025