Access to this page requires authorization. You can try signing in or changing directories . Access to this page requires authorization. You can try changing directories . Currently viewing: Foundry (classic) portal version - Switch to version for the new Foundry portal GPT Realtime Translate is a purpose-built model for continuous, real-time audio translation. Unlike pipeline-based approaches that chunk audio into segments before translating, this model processes audio streams continuously and produces translated output as speech unfolds. Use it in scenarios where audio is flowing live and latency must be minimal. Key capabilities Continuous stream processing : Translates live audio without segmenting or buffering, producing output that tracks the cadence of the original speech. Speech and text output : Produces both translated speech (audio) and a translated transcript in the target language. Low-latency translation : Keeps pace with real-time conversation, reducing the gap between the original speech and translated output. When to use GPT Realtime Translate Use GPT Realtime Translate when you need: Live streaming events, conferences, and broadcasts requiring real-time multilingual output. Cross-language customer support calls. Multilingual voice interfaces and applications. Live media localization. International real-time meetings and collaboration. Example use cases Live multilingual events : Translate conference talks, webinars, or broadcasts in real time so audiences can listen in their preferred language. Pair with GPT Realtime Transcribe to simultaneously provide source-language captions. Global customer support : Route inbound calls through GPT Realtime Translate to bridge language gaps between customers and agents. The translated transcript gives agents a written record in their language for follow-up.

International voice assistants : Build once and deploy across languages. GPT Realtime Translate enables multilingual voice interactions without requiring per-language model deployments. GPT Realtime Translate is available through the Realtime API. The connection and usage patterns are the same as for other realtime models: Use the GPT Realtime API via WebSockets Use the same quickstart code patterns, and set your deployment name to the model deployment you created for gpt-realtime-translate . Language support guidance GPT Realtime Translate is designed for multilingual scenarios. Translation quality can vary by language pair, domain vocabulary, speaking style, and audio quality. If you enable transcription settings in your session, provide an ISO-639-1 language hint (for example, en ) when available. Validate your target language pairs with production-like audio before rollout. For broader language and locale references, see Language and voice support for the Speech service . Deployment and availability GPT Realtime Translate is available as a Global Standard (pay-as-you-go) deployment in Microsoft Foundry. Deploy the model from the model catalog . For current region support, see Region availability for Foundry Models sold by Azure . gpt-realtime-translate is priced hourly. For current rates, see the Audio Models section on the Azure OpenAI pricing page . Was this page helpful? Need help with this topic? Want to try using Ask Learn to clarify or guide you through this topic? Additional resources Last updated on 2026-06-05