XFINITY / TV VOICE EXPERIENCE
Designing the visual language of voice on TV.
Customers pressed the microphone button on an Xfinity Voice Remote to search, navigate, and control TV by speaking. But the on-screen experience had to make an invisible interaction understandable: what the system heard, what customers could say next, and how to act on a response. I was a design lead on the evolution of that experience, shaping feedback, contextual guidance, and richer responses.
Interaction design · Multimodal UX · Contextual education

Voice Tips: short, contextual prompts that introduced useful voice commands during everyday TV interactions. Here, a tip suggests saying “Resume” while browsing a show.
01 / OVERVIEW
This wasn't the original launch of voice control. It was a major evolution of an established product: bringing individual voice interactions into a clearer, more consistent TV experience while preserving the context of what people were watching.
ROLE
Associate Creative Director and hands-on design lead. I led Voice Tips, advocated for a consistent and less intrusive on-screen feedback strategy, and contributed to interaction and visual decisions across the wider TV voice experience. I worked closely with UX, UI, content, research, and product partners.
SCOPE
Interaction and visual design · Voice education · Experience patterns · Cross-functional leadership
THE BIG PICTURE
WHAT THIS WORK DEMONSTRATES
Design decisions at different scales.
These examples show how individual interaction decisions connected to customer research, reusable patterns, collaborative leadership, and measurable product adoption.
Hands-on interaction and visual design
Designed and directed recognizable treatments for listening, live recognition, spoken commands, and contextual guidance.
Research-informed product decisions
Used usage patterns and customer insights to focus education on moments when people could immediately use a new command.
Reusable interface and content patterns
Connected visual feedback, command wording, and on-screen placements so related interactions felt like one system.
Cross-functional creative leadership
Advocated for a more consistent on-screen approach while leading and contributing to design decisions across UX, UI, research, and content.
Real-world adoption and measurable results
Voice Tips shipped, followed by a reported 45% increase in voice-command usage.
02 / DESIGN CHALLENGE
How do you give an invisible interaction a visible language?
A spoken command doesn't show its own status or next step. Viewers needed to know when TV was listening, whether it understood, how to discover commands, and how to handle answers that weren't straightforward. The design challenge was making those moments clear without disrupting the entertainment experience.
Make behavior legible
Differentiate listening, live speech recognition, and completed input with unmistakable on-screen states.
Make capabilities discoverable
Teach a useful command in context, with the exact words to say visually distinguished.
Make responses actionable
Resolve similar search results with distinguishing metadata, numbered voice choices, and remote navigation.
03 / MAKE VOICE UNDERSTANDABLE
An invisible input needed recognizable feedback.
We developed distinct feedback for active listening, live speech recognition, completed speech, and contextual guidance. I advocated for giving these lightweight interactions a consistent, compact location rather than a large, screen-dominating treatment. The bigger decision was to make system status recognizable while letting the content remain the focus.

The new lightweight feedback treatment moved from the lower screen to a consistent upper-left location.

A consistent microphone treatment uses ring states and concise text to communicate listening, word-by-word recognition, completed speech, errors, and contextual suggestions.
04 / MAKE CAPABILITIES DISCOVERABLE
Teaching customers what to say, when it mattered.
We introduced Voice Tips, contextual prompts that helped customers discover commands they might not otherwise know. Rather than relying on a separate help experience, we integrated suggestions into familiar moments like browsing, searching, and controlling playback. I led the design of this feature, connecting the on-screen interaction, content approach, and visual treatment.
Why contextual education mattered.
August 2018 usage data indicated that roughly 90% of voice-remote households were light or moderate users, with fewer than about 100 commands per month. That suggested an opportunity to teach one useful behavior at a time, in the moment when someone could try it. We focused on activities such as browsing, search, and launching apps.

A voice shortcut appears alongside an action the person already understands.
Designing the suggestion as a repeatable pattern.
With UX, UI, research, and content partners, we designed a repeatable treatment that separated conversational guidance from the exact words to say. The command appeared at the end, emphasized in bold italic type, so the screen could both explain the opportunity and give customers a phrase they could immediately repeat.

The suggestion blueprint connects visual treatment and content structure. Select any artifact to view it larger.
Related work: “What can I say?”
Voice Tips provided in-context education. In a related initiative, I served as art director on a redesign of “What Can I Say?”, working with a dedicated design lead and research and content partners on a more engaging, full-screen way to explore commands. That separate experiment reported a 20–70% increase in usage of the commands it promoted.
05 / MAKE RESPONSES MORE USEFUL
Make the next action clear, even when the answer isn't.
Voice responses also needed to help customers move forward. Two related design challenges illustrate that work: resolving ambiguous entertainment searches without repeated commands, and making informational responses useful through follow-up actions. I contributed to the experience and design direction across this work alongside UX and UI partners.
A title isn't always enough to identify the right result.
A spoken search for “Beauty and the Beast” could match several titles. We distinguished the most relevant results with format and release information, then numbered the choices so viewers could select one by speaking a number or using the remote. The interaction accommodated ambiguity without requiring a more complicated voice command.
EXAMPLE VOICE QUERY / “BEAUTY AND THE BEAST”

Distinguishing labels, release years, and numbered results helped people choose among similarly named titles by voice or remote.
Going beyond answers with interactive Voice Cards.
In related 2019 work, we introduced “pivots” so customers could act on information shown in a lower-third response. After identifying a song, for example, a viewer could open artist information, search YouTube, or launch a music partner without restarting the interaction. The UX lead defined pivot behaviors and phased them with engineering; I contributed to design direction with the UX and UI team.

Voice Cards added pivots such as artist information and music-provider actions so a spoken answer could lead directly to the next task.
One visual language, multiple interactions.
Across listening feedback, contextual suggestions, and richer responses, consistent color, typography, and on-screen treatments helped the experience feel coherent. I helped shape that shared design direction. These style-guide excerpts capture the foundational color and typography guidance for designers.


Visual style-guide excerpts: semantic color roles and typography.
06 / ADOPTION + OUTCOME
45%
Increase in voice-command usage after Voice Tips launched.
The discovery feature was followed by approximately 5% year-over-year growth.
Feature outcome
Voice Tips made new capabilities discoverable in moments when people could immediately try them. The reported increase in voice-command usage after launch provides the clearest measurable outcome for this part of the wider redesign.
Broader design impact
The broader work gave feedback, guidance, search, and richer responses a more coherent interaction language. Years later, I applied related principles to intelligent-assistant behavior across TV and mobile.