gexgd0419/NaturalVoiceSAPIAdapter

NaturalVoiceSAPIAdapter – a Windows SAPI 5 TTS engine that taps Microsoft’s natural‑voice services

What it is

• A drop‑in SAPI 5 text‑to‑speech (TTS) engine for Windows that lets any program that uses SAPI 5 voices speak with the high‑quality natural (neural) voices Microsoft provides.
• It works with three sources of voices:
  1. Local Narrator voices that ship with Windows 11 (and can be made to run on older Windows versions via a hack).
  2. Online voices used by Microsoft Edge’s Read Aloud feature.
  3. Azure AI Speech Service voices – you supply your own Azure subscription key.

Why it matters

• Most Windows TTS engines are limited to the older, robotic‑sounding system voices. This adapter unlocks the modern, human‑like neural voices for any SAPI‑compatible application (e.g., screen readers, voice‑over tools, custom software).
• It bridges the gap between desktop applications and cloud‑based speech services without requiring each app to implement its own Azure client.

Key features

• 32‑bit and 64‑bit versions (install both on a 64‑bit OS to support all programs).
• Automatic registration of the engine via a simple GUI installer (Installer.exe).
• Optional toggles for:
  • Enabling/disabling Edge online voices.
  • Enabling/disabling Azure online voices (requires API key and region).
• Language filter – by default only the languages you have set in the installer are exposed, avoiding a cluttered voice list.
• Registry‑based configuration for advanced tweaks (see the wiki link for the list of values).
• Test utility (TtsApplication.exe) that shows phoneme/viseme events and supports Chinese UI.

System requirements

• Tested on Windows XP SP3 (32‑bit) up through Windows 11.
• Local Narrator natural voices need at least Windows 7; they can be installed on Windows 10 (build 17763+) via provided MSIX packages.
• Internet connection is required for any online voice source.
• For Azure voices you need a valid Azure Speech Service subscription key.

Installation steps

1. Download the zip from the Releases page.
2. Extract it without moving or renaming the files.
3. Run Installer.exe (requires admin rights).
4. Choose which voice sources to enable and, if using Azure, enter your key/region.
5. Click Install; the installer will register the appropriate 32‑/64‑bit DLLs.
6. (Optional) Register the DLLs manually with regsvr32.

Using the engine

• After installation, any Windows program that lets you pick a SAPI 5 voice will list the new “NaturalVoice” entries.
• You can also open the classic Control Panel → Speech dialog (or Speech Recognition → Text to Speech on Vista+). The installed voices appear there.
• For quick testing, launch x86\TtsApplication.exe or x64\TtsApplication.exe.

Technical notes

• The engine extracts encryption keys from system files to unlock the Narrator/Edge voices – essentially a hack, so future Windows updates could break it.
• It relies on several open‑source libraries: websocketpp, standalone ASIO, OpenSSL, nlohmann/json, YY‑Thunks (for XP compatibility), and spdlog for logging.
• The core speech synthesis is powered by Microsoft’s Microsoft.CognitiveServices.Speech.Extension.Embedded.TTS library.

Where to find more

• Detailed technical info, registry options, and voice‑download links are in the repository’s wiki.
• The Chinese README (README.zh.md) mirrors this information for Mandarin speakers.

---

NaturalVoiceSAPIAdapter is a practical tool for developers and power users who need high‑quality neural TTS on Windows without rewriting their applications for Azure or Edge APIs.