Google Meet AI Voice Translation
A concept and research project designing real-time AI voice translation natively into Google Meet, so people across different languages can stay in the conversation.
MY ROLE
UX & Interaction Designer
MY TEAM
TIMELINE
3 Months
5
Research Participants
3
Core Features Designed
I/O 2025
Google validated this project
PROBLEM
Google Meet connects people across the world, but when conversations cross languages, the experience breaks down, users fall behind, lose context, and reach for Google Translate mid-call.
OUTCOME
Designed a privacy-first translation feature with live captions and post-call transcripts, built natively into Meet's interface, and prototyped it with real AI-generated voice and video to make it feel real.
Google Meet AI Voice Translation
A concept and research project designing real-time AI voice translation natively into Google Meet, so people across different languages can stay in the conversation.
MY ROLE
UX & Interaction Designer
MY TEAM
TIMELINE
3 Months
PROBLEM
Google Meet connects people across the world, but when conversations cross languages, the experience breaks down, users fall behind, lose context, and reach for Google Translate mid-call.
OUTCOME
Designed a privacy-first translation feature with live captions and post-call transcripts, built natively into Meet's interface, and prototyped it with real AI-generated voice and video to make it feel real.
5
Research Participants
3
Core Features Designed
I/O 2025
Google validated this project
Google Meet AI Voice Translation
A concept and research project designing real-time AI voice translation natively into Google Meet, so people across different languages can stay in the conversation.
MY ROLE
UX & Interaction Designer
MY TEAM
TIMELINE
3 Months
PROBLEM
Google Meet connects people across the world, but when conversations cross languages, the experience breaks down, users fall behind, lose context, and reach for Google Translate mid-call.
OUTCOME
Designed a privacy-first translation feature with live captions and post-call transcripts, built natively into Meet's interface, and prototyped it with real AI-generated voice and video to make it feel real.
5
Research Participants
3
Core Features Designed
I/O 2025
Google validated this project
Context
Context
This project was designed to live inside Google Meet, not as a standalone translation tool, but as a native feature that fits into how a call already works, built for the moment language becomes a barrier.
This project was designed to live inside Google Meet, not as a standalone translation tool, but as a native feature that fits into how a call already works, built for the moment language becomes a barrier.
Stake
Design a translation feature people would actually trust and use mid-call.
Constraint
Concept project with no access to Google's design system or internal usage data.
Success
A feature that feels native to Meet, not bolted on.
Why Now?
Why Now?
Remote work made multilingual meetings the norm, but the tools never caught up.
Remote work made multilingual meetings the norm, but the tools never caught up.
1.8B Google Meet Users
1.8B Google Meet Users
1.8B+ people use Google Meet globally across work, education, and healthcare
1.8B+ people use Google Meet globally across work, education, and healthcare
Non-native English Speakers
Non-native English Speakers
Non-native English speakers make up the majority of Google Meet's user base
Non-native English speakers make up the majority of Google Meet's user base
Google Translate Workaround
Google Translate Workaround
Existing workarounds like Google Translate require leaving the call entirely
Existing workarounds like Google Translate require leaving the call entirely
Real-time AI translation
Real-time AI translation
Real-time AI translation became technically feasible, but no platform had integrated it
Real-time AI translation became technically feasible, but no platform had integrated it
Stake
Design a translation feature people would actually trust and use mid-call.
Constraint
Concept project with no access to Google's design system or internal usage data.
Success
A feature that feels native to Meet, not bolted on.
User Research
User Research
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
01 Community Signals
01 Community Signals
5 participants across 5 sessions
5 participants across 5 sessions
One participant stayed constant; the other changed each time, representing a different language or accent. Actions, reactions, pain points, and frustration recorded in real time.
One participant stayed constant; the other changed each time, representing a different language or accent. Actions, reactions, pain points, and frustration recorded in real time.
02 Concept Testing
02 Concept Testing
All 5 participants responded after seeing the feature Introduced the translation concept post-session and collected direct feedback on expectations, concerns, and placement preferences.
All 5 participants responded after seeing the feature Introduced the translation concept post-session and collected direct feedback on expectations, concerns, and placement preferences.
03 HSE Principles Framework
03 HSE Principles Framework
5 principles applied: Perception, Attention, Cognitive Load, Expectations, User Control
Mapped observations to design decisions, turning behavioral patterns into structured design rationale.
5 principles applied: Perception, Attention, Cognitive Load, Expectations, User Control
Mapped observations to design decisions, turning behavioral patterns into structured design rationale.
04 Competitive Landscape Tools analysed
04 Competitive Landscape Tools analysed
Google Translate, Google Meet captions
Every workaround required leaving the call. No platform had natively integrated real-time voice translation, confirming a clear product gap.
Google Translate, Google Meet captions
Every workaround required leaving the call. No platform had natively integrated real-time voice translation, confirming a clear product gap.
Google Meet Captions
Google Meet Captions
Google Translation
Google Translation
Final Research Impact: Every feature, consent first, live captions, post-call transcript, mapped directly to what participants said they needed.
Final Research Impact: Every feature, consent first, live captions, post-call transcript, mapped directly to what participants said they needed.
Problem & Opportunity
Problem & Opportunity
Google Meet connects the world, but when conversations cross languages, the experience breaks, users lose context, fall behind, and reach for workarounds that pull them out of the call.
Google Meet connects the world, but when conversations cross languages, the experience breaks, users lose context, fall behind, and reach for workarounds that pull them out of the call.
Before
Before
Google Meet = Captions only, no translation
Google Meet = Captions only, no translation
Missing
Missing
Native, real-time voice translation with user control
Native, real-time voice translation with user control
Opportunity
Opportunity
A privacy-first translation feature built into the call itself
A privacy-first translation feature built into the call itself
Target User
Target User
People in multilingual meetings who need to communicate across language barriers without disrupting the flow of the call.
People in multilingual meetings who need to communicate across language barriers without disrupting the flow of the call.
International Student
International Student
Remote-team Collaborator
Remote-team Collaborator
Healthcare Professionals
Healthcare Professionals
Cross-border Business User
Cross-border Business User
User Core Needs
User Core Needs
Minimal Delay
Minimal
Delay
Accurate Translation
Accurate Translation
Privacy Control
Privacy
Control
One-click Access
One-click Access
Post-call clarity
Post-call Clarity
Product Goals & Strategy
Product Goals & Strategy
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
The tradeoff we optimised for:
The tradeoff we optimised for:
Trade Off
Opt in > Always On
Defaulting to off builds trust faster than defaulting to on.
Fewer controls, placed better, beat more controls buried deeper.
Native UI > Feature Rich
Explanation
Trade Off
Opt in > Always On
Defaulting to off builds trust faster than defaulting to on.
Fewer controls, placed better, beat more controls buried deeper.
Native UI > Feature Rich
Explanation
Ideation
Ideation
One clear direction from the start: build translation natively into Meet, not around it.
One clear direction from the start: build translation natively into Meet, not around it.
Product Direction
Product Direction
Real-time AI voice translation
Real-time AI voice translation
Live
Captions
Live
Captions
Consent Upfront
Consent Upfront
Post-call trranscript
Post-call trranscript
+
+
+
Final Plan
Final Plan
Research
Insights
Research
Insights
Consent
Design
Consent
Design
Live
Translation
Live
Translation
Post-call
transcript
Prototype
Post-call
transcript
Prototype
Design System & Visual Language
Design System & Visual Language
Since Google Meet lives within the Google Ecosystem, I worked within Material 3 Guidelines, using Google Sans typography, and Google's color system to ensure every feature felt like it already belonged.
Since Google Meet lives within the Google Ecosystem, I worked within Material 3 Guidelines, using Google Sans typography, and Google's color system to ensure every feature felt like it already belonged.
Addition to Design System
Addition to Design System
Speaker and Listen Dropdowns
Current Google Meet Icons
AI Speaking Indicator
Request Participants Button
Live Transcript Labels
Voice Translation Toggle
Speaker and Listen Dropdowns
Current Google Meet Icons
AI Speaking Indicator
Request Participants Button
Live Transcript Labels
Voice Translation Toggle
Key Screens
Key Screens
Primary Consent
Permission Request
Shows exactly who requested translation and why, before anything is recorded
Language Pair
Speaker and listener languages set upfront, no surprises mid-call
Approve Request CTA
Nothing starts until the user actively confirms
Permission Request
Shows exactly who requested translation and why, before anything is recorded
Language Pair
Speaker and listener languages set upfront, no surprises mid-call
Approve Request CTA
Nothing starts until the user actively confirms
Live Translation
Hear Translation Toggle
One-click control to turn audio translation on or off without leaving the come all
Live Dual-Language Transcript
Both the original and translated text shown simultaneously, inline
Active Speaker Indicator
Blue ring signals who is currently being translated
Hear Translation Toggle
One-click control to turn audio translation on or off without leaving the come all
Live Dual-Language Transcript
Both the original and translated text shown simultaneously, inline
Active Speaker Indicator
Blue ring signals who is currently being translated
Post-Call Transcript
Translation Summary Card
Surfaces automatically after the call, no searching for it
Privacy Reminder
Confirms transcript complies with privacy policies, reinforcing trust post-call
Download Transcript
Full bilingual transcript available immediately
Translation Summary Card
Surfaces automatically after the call, no searching for it
Privacy Reminder
Confirms transcript complies with privacy policies, reinforcing trust post-call
Download Transcript
Full bilingual transcript available immediately
Outcomes and Ownership
Outcomes and Ownership
Outcomes
Outcomes
A concept and research project that went from structured user observation to a fully functioning AI voice prototype, validating a real product gap before Google shipped the same idea.
A concept and research project that went from structured user observation to a fully functioning AI voice prototype, validating a real product gap before Google shipped the same idea.
5
5
Research Participants
Research Participants
3
3
Core Features Designed
Core Features Designed
I/O 2025
I/O 2025
Google Validated the concept
Google Validated the concept
Ownership
Ownership
I led the end-to-end design, from research synthesis to interaction design, component decisions, and prototype production using Veo3 and ElevenLabs.
I led the end-to-end design, from research synthesis to interaction design, component decisions, and prototype production using Veo3 and ElevenLabs.
Ownership Flow
Ownership Flow
Research
Prototype
HSE
HSE
Framework
Framework
Research
Interaction
Interaction
Design
Design
System
System
Design
Design
Prototypes
Validation
Validation
Context
This project was designed to live inside Google Meet, not as a standalone translation tool, but as a native feature that fits into how a call already works, built for the moment language becomes a barrier.
Why Now?
Remote work made multilingual meetings the norm, but the tools never caught up.
1.8B Google Meet Users
1.8B+ people use Google Meet globally across work, education, and healthcare
Non-native English Speakers
Non-native English speakers make up the majority of Google Meet's user base
Google Translate Workaround
Existing workarounds like Google Translate require leaving the call entirely
Real-time AI translation
Real-time AI translation became technically feasible, but no platform had integrated it
Stake
Design a translation feature people would actually trust and use mid-call.
Constraint
Concept project with no access to Google's design system or internal usage data.
Success
A feature that feels native to Meet, not bolted on.
Problem & Opportunity
Google Meet connects the world, but when conversations cross languages, the experience breaks, users lose context, fall behind, and reach for workarounds that pull them out of the call.
Before
Google Meet = Captions only, no translation
Missing
Native, real-time voice translation with user control
Opportunity
A privacy-first translation feature built into the call itself
Target User
People in multilingual meetings who need to communicate across language barriers without disrupting the flow of the call.
International Student
Remote-team Collaborator
Healthcare Professionals
Cross-border Business User
User Core Needs
Minimal Delay
Minimal Delay
Accurate
Translation
Accurate
Translation
Provacy Control
Provacy Control
One-click
Access
One-click
Access
Post-call
Clarity
Post-call
Clarity
Ideation
One clear direction from the start: build translation natively into Meet, not around it.
Final Plan
Research Insights
Consent Design
Live Translation
Post-call transcript
Prototype
Design System & Visual Language
Since Google Meet lives within the Google Ecosystem, I worked within Material 3 Guidelines, using Google Sans typography, and Google's color system to ensure every feature felt like it already belonged.
Addition to Design System
See you next time around!
Thank You
GET IN TOUCH
I'm currently open for exciting opportunities.
Let’s connect and talk about the next big thing!
dhvans © 2025
Designed with 🤍 by Dhvani Shah
Thank You
GET IN TOUCH
See you next time around!
I'm currently open for exciting opportunities.
Let’s connect and talk about the next big thing!
dhvans © 2025
Designed with 🤍 by Dhvani Shah
See you next time around!
Thank You
GET IN TOUCH
I'm currently open for exciting opportunities.
Let’s connect and talk about the next big thing!
dhvans © 2025
Designed with 🤍 by Dhvani Shah
Final Research Impact: Every feature, consent first, live captions, post-call transcript, mapped directly to what participants said they needed.
User Research
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
01 Community Signals
5 participants across 5 sessions
People were already trying to stay accountable together through shared presence, check-ins, and informal co-working.
02 Direct User Research
All 5 participants responded after seeing the feature Introduced the translation concept post-session and collected direct feedback on expectations, concerns, and placement preferences.
03 HSE Principles Framework
5 principles applied: Perception, Attention, Cognitive Load, Expectations, User Control
Mapped observations to design decisions, turning behavioral patterns into structured design rationale.
04 Competitive Landscape Tools analysed
Google Translate, Google Meet captions
Every workaround required leaving the call. No platform had natively integrated real-time voice translation, confirming a clear product gap.
Google Meet Captions
Google Translate
Product Goals & Strategy
Structured observation sessions and concept testing to understand where language barriers break down in live calls, and what users actually needed from a translation feature.
The tradeoff we optimised for:
Outcomes
A concept and research project that went from structured user observation to a fully functioning AI voice prototype, validating a real product gap before Google shipped the same idea.
5
Research Participants
3
Core Features Designed
I/O 2025
Google Validated the concept
Outcomes & Ownership
Ownership
I led the end-to-end design, from research synthesis to interaction design, component decisions, and prototype production using Veo3 and ElevenLabs.
Ownership Flow
HSE Framework
Interaction Design
Research
Prototype
System Design
Validation
Key Screens
Primary Consent
Live Translation
Post-Call Transcript
