Azure AI Speech (Neural TTS)

Azure AI Speech's text-to-speech service (formerly Cognitive Services Speech, now part of Azure AI Foundry Tools) converts text into natural-sounding synthesized speech using deep neural networks. The standard catalog covers 100+ languages and locales with both 24 kHz and 48 kHz high-fidelity output, plus newer HD voices and Azure OpenAI voices for higher-quality scenarios. Beyond out-of-the-box voices, it offers professional voice fine-tuning and 'personal voice' cloning for branded experiences, plus a text-to-speech avatar feature that renders synced video.

It's aimed at developers and enterprises that need a production TTS engine wired into a broader cloud: real-time synthesis via SDK or REST, asynchronous batch synthesis for long-form audio like audiobooks, SSML for pitch/pause/pronunciation control, and viseme events for lip-sync animation. Billing is pay-as-you-go per character (Chinese, Japanese kanji, and Korean hanja count as two characters each), with separate per-hour pricing for custom voice training and endpoint hosting, and per-second billing for avatar video.

Custom Neural Voice is gated behind an access application as part of Microsoft's responsible-AI controls, and SSML support varies by voice family (HD, personal, and embedded voices don't accept the full tag set). Integration is straightforward if you're already on Azure, with Speech Studio offering a no-code Audio Content Creation tool and sample code across most major languages.