AI UX PlaygroundNewsletterJoin 2K+ AI designers and PMs on Substack. New teardowns, patterns, and prompts as they drop.

Audio

Real-time Translation

Transcribe and translate spoken conversation live, with speaker labels when it helps. Latency and accuracy matter more than pretty UI chrome.

Interactive demo

Real-time translation

Input (English)

Hello, how are you?

Overview

How might we design real-time translation so people can trust and act on AI output?

When to use

  • Perfect for international communication tools, customer support platforms, and collaboration applications where real-time language translation enables seamless multilingual conversations.

When to skip

  • Legal testimony requiring certified human interpreters.
  • Audio so noisy that live MT error rates break trust.
  • Languages your model pair does not support reliably.

Rules

  • Translation with no original language visible.

  • Lag of many seconds with no partial transcript.

  • Wrong speaker attribution in multi-party rooms.

  • No mute or privacy notice while mic is live.

Evidence

ProductImplementation
Google TranslateConversation mode with split-screen bilingual text.
Microsoft TranslatorLive captions and translation in Teams meetings.
ZoomTranslated captions during webinars and calls.
SkypeReal-time call translation with on-screen text.

FAQ

What should live translation UI show?

Original and translated text, active speaker, and language pair. Partials update as speech continues.

Voice or text output?

Captions are default. Optional TTS playback for hands-free, with clear latency tradeoffs.

How handle domain jargon?

Custom glossaries or user corrections that persist for the session.

Live translation vs lang-toggle?

Live translation is spoken input in meetings. Lang-toggle switches output language of written AI content.