Microsoft Azure AI Speech, now branded Azure Speech in Foundry Tools, is a cloud service for adding voice to applications and agents. It covers speech to text, text to speech, real-time speech translation and speaker recognition. Developers can build custom neural voices, create prebuilt or custom avatars, transcribe with the Whisper model, or connect voice to AI agents through the Voice Live API. Billing is pay-as-you-go by audio hours transcribed, characters converted to speech and speaker recognition transactions, with no upfront costs.
Software developers, AI engineers and enterprise teams are the main users, especially in call centers, contact centers and customer-facing apps. Product teams that need multilingual captions, brand-specific voices or voice-enabled agents can start from prebuilt models and customize them. Organizations with strict data or connectivity requirements can run models in containers or use embedded speech on devices, and regulated industries can rely on Microsoft's security and compliance certifications.
Common scenarios include transcribing call center or meeting conversations, captioning audio in more than 100 languages, and building bots that speak naturally in a distinctive voice. Post-call analytics use foundation models in Azure Content Understanding to analyze recorded calls, while translation turns audio or video into other languages for multilingual communication. Voice agents can pair transcription, a custom voice and an avatar in a single experience, and the service works alongside Azure OpenAI, Microsoft Foundry, Azure Translator and Azure Language. Embedded speech keeps speech to text and text to speech running where connectivity is intermittent or unavailable.
