MultiSpeaker – Real-time Multilingual Voice Translator

MultiSpeaker is a real-time multilingual translation platform built on top of Azure Speech Services. The system captures live audio from the user’s microphone, detects speech segments and streams them to Azure for speech-to-text and text-to-text translation. For each configured target language, MultiSpeaker generates both a translated text transcript and a natural-sounding neural voice output. In practice, this turns a single spoken language into many parallel channels: original speech, translated subtitles and translated audio. Participants can listen in their preferred language while still following the original speaker. This makes MultiSpeaker useful for distributed teams, international workshops, hybrid university classes and public events where language is a barrier. The interface is organized around sessions. In each session you can select input language, choose one or more output languages and configure which voices to use per language. As people speak, the UI shows per-language captions in real time and, if enabled, plays back translated audio. Under the hood, MultiSpeaker uses Azure’s speech recognition, translation and text-to-speech APIs, combined with a low-latency streaming backend to keep delays as small as possible. Our focus is on reliability, clarity and accessibility: clear transcripts, readable subtitles and stable audio output. As the project evolves, MultiSpeaker can be extended with speaker labelling, per-user language preferences, recording/exports and integrations into existing meeting tools – but the core idea remains simple: one spoken voice, many languages, in real time.