Introducing updated GPT Voice Models in Microsoft Foundry

2025-12-17 · Source: Microsoft Foundry Blog articles · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

Microsoft Foundry has released updated GPT voice models, now generally available via API, designed for production-ready real-time audio applications. These include "gpt-realtime-mini-2025-12-15" for low-latency voice agents with improved prosody and voice fidelity, supporting voice cloning for trusted customers. The "gpt-4o-mini-transcribe-2025-12-15" model offers up to 50% lower word error rate (WER) on English benchmarks and reduces hallucinations on silence by up to 4x. Additionally, "gpt-4o-mini-tts-2025-12-15" provides multilingual speech synthesis with 35% fewer word errors on multilingual benchmarks and up to 3x lower WER in non-English languages, also supporting voice cloning. All these enhancements are available without any changes to current pricing.

Key takeaway

For AI Architects and NLP Engineers building real-time voice applications, these updated GPT models in Microsoft Foundry offer significant performance gains in accuracy, latency, and multilingual support without increased cost. You should evaluate "gpt-realtime-mini-2025-12-15" for voice agents, "gpt-4o-mini-transcribe-2025-12-15" for transcription, and "gpt-4o-mini-tts-2025-12-15" for speech synthesis to enhance your systems and maintain brand voice consistency through features like voice cloning.

Key insights

Microsoft Foundry's new GPT voice models offer enhanced real-time performance, accuracy, and multilingual capabilities at existing price points.

Principles

• Reliability and latency are critical for production voice models.
• Voice cloning ensures consistent brand voice.
• Cost-efficiency is maintained despite performance upgrades.

In practice

• Use "gpt-realtime-mini" for low-latency voice agents.
• Employ "gpt-4o-mini-transcribe" for robust real-time transcription.
• Apply "gpt-4o-mini-tts" for natural multilingual speech synthesis.

Topics

• GPT Voice Models
• Speech Synthesis
• Speech Transcription
• Voice Cloning
• Microsoft Foundry API

Best for: AI Architect, NLP Engineer, CTO, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• A New Chapter for Realtime AI: Reasoning, Translation, and Real-Time Transcription — New OpenAI models enhance real-time AI voice applications with advanced translation, transcription, and reasoning capabilities.
• Introducing MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 in Microsoft Foundry — Microsoft Foundry introduces three new AI models for efficient, high-quality speech recognition, speech generation, and text-to-image capabilities.
• GPT-5.6 Pro LEAKED & Is Coming Soon! Mythos 5 Level! — GPT 5.6 Pro elevates LLM performance through enhanced reasoning, integrated tooling, and sophisticated code and simulation generation.
• How Mistral’s Voxtral TTS is Redefining Multilingual Voice Cloning with a Hybrid Autoregressive and Flow-Matching Architecture — Voxtral TTS uses a hybrid autoregressive and flow-matching architecture for high-fidelity multilingual voice cloning.
• Angelos Perivolaropoulos of ElevenLabs at RAAIS 2026 — Real-world voice AI demands balancing high speech-to-text accuracy with critical low-latency performance for effective interactive systems.
• OpenAI's NEW Voice Agent Model - GPT-RealTime 2 is dope! — OpenAI's new real-time voice models offer low-latency, bidirectional voice communication, translation, and transcription.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Microsoft Foundry Blog articles.