SaschaCowley · Jan 23, 2025 · Jan 8, 2025 · Jan 8, 2025 · Jan 8, 2025 · Jan 8, 2025
diff --git a/source/config/configSpec.py b/source/config/configSpec.py
@@ -45,6 +45,7 @@
 	autoDialectSwitching = boolean(default=false)
 	delayedCharacterDescriptions = boolean(default=false)
 	excludedSpeechModes = int_list(default=list())
+	hasSapi4WarningBeenShown = boolean(default=False)
 
 	[[__many__]]
 		capPitchChange = integer(default=30,min=-100,max=100)

diff --git a/source/speech/speech.py b/source/speech/speech.py
@@ -11,14 +11,17 @@
 import weakref
 import unicodedata
 import time
+
 import colors
 import api
 from annotation import _AnnotationRolesT
 import controlTypes
 from controlTypes import OutputReason, TextPosition
 from controlTypes.state import State
+from gui.message import MessageDialog
+import queueHandler
 import tones
-from synthDriverHandler import getSynth
+from synthDriverHandler import SynthDriver, getSynth, synthChanged
 import re
 import textInfos
 import speechDictHandler
@@ -3059,3 +3062,34 @@ def clearTypedWordBuffer() -> None:
 	complete the word (such as a focus change or choosing to move the caret).
 	"""
 	_curWordChars.clear()
+
+
+def _sapi4DeprecationWarning(synth: SynthDriver, audioOutputDevice: str, isFallback: bool):
+	"""A synthChanged event handler to alert the user about the deprecation of SAPI4."""
+
+	def setShown():
+		config.conf["speech"]["hasSapi4WarningBeenShown"] = True
+
+	def impl():
+		MessageDialog(
+			parent=None,
+			message="Microsoft Speech API version 4 is obsolete. "
+			"Using this speech synthesizer may pose a security risk. "
+			"This synthesizer driver will be removed in NVDA 2026.1. "
+			"You are strongly encouraged to choose a more modern speech synthesizer.",
+			title="Warning",
+			buttons=None,
+		).addOkButton(
+			callback=setShown,
+		).Show()
+
+	if (
+		(not isFallback)
+		and (synth.name == "sapi4")
+		and (not config.conf["speech"]["hasSapi4WarningBeenShown"])
+	):
+		# We need to queue the dialog to appear, as wx may not have been initialised the first time this is called.
+		queueHandler.queueFunction(queueHandler.eventQueue, impl)
+
+
+synthChanged.register(_sapi4DeprecationWarning)
@@ -1,7 +1,5 @@
-# _sapi4.py
-# Contributed by Serotek Corporation under the GPL
 # A part of NonVisual Desktop Access (NVDA)
-# Copyright (C) 2006-2008 NVDA Contributors <http://www.nvda-project.org/>
+# Copyright (C) 2006-2025 NV Access Limited, Serotek Corporation
 # This file is covered by the GNU General Public License.
 # See the file COPYING for more details.
 
@@ -18,6 +16,7 @@
 	POINTER,
 	sizeof,
 	Structure,
+	windll,
 )
 from ctypes.wintypes import BYTE, DWORD, LPCWSTR, WORD
 from comtypes import GUID, IUnknown, STDMETHOD
@@ -227,3 +226,21 @@ class ITTSNotifySinkW(IUnknown):
 
 CLSID_MMAudioDest = GUID("{CB96B400-C743-11cd-80E5-00AA003E4B50}")
 CLSID_TTSEnumerator = GUID("{D67C0280-C743-11cd-80E5-00AA003E4B50}")
+
+
+# WaveOutMessage message codes
+# Defined in mmddk.h
+DRV_QUERYFUNCTIONINSTANCEID = 2065
+DRV_QUERYFUNCTIONINSTANCEIDSIZE = 2066
+# Defined in mmsyscom.h
+MMSYSERR_NOERROR = 0
+
+# Function prototypes
+# Defined in mmeapi.h
+winmm = windll.winmm
+waveOutMessage = winmm.waveOutMessage
+waveOutMessage.restype = c_uint
+
+waveOutGetNumDevs = winmm.waveOutGetNumDevs
+waveOutGetNumDevs.argtypes = []
+waveOutGetNumDevs.restype = c_int
@@ -1,18 +1,19 @@
 # A part of NonVisual Desktop Access (NVDA)
 # Copyright (C) 2006-2024 NV Access Limited, Leonard de Ruijter
 # This file is covered by the GNU General Public License.
-# See the file COPYING for more details.
+# See the file COPYING for  Not e details.
 
 import locale
 from collections import OrderedDict
 import winreg
 from comtypes import CoCreateInstance, COMObject, COMError, GUID
-from ctypes import byref, c_ulong, POINTER
-from ctypes.wintypes import DWORD, WORD
+from ctypes import byref, c_ulong, POINTER, c_wchar, create_string_buffer, sizeof
+from ctypes.wintypes import DWORD, HANDLE, WORD
 from typing import Optional
 from synthDriverHandler import SynthDriver, VoiceInfo, synthIndexReached, synthDoneSpeaking
 from logHandler import log
 from ._sapi4 import (
+	MMSYSERR_NOERROR,
 	CLSID_MMAudioDest,
 	CLSID_TTSEnumerator,
 	IAudioMultiMediaDevice,
@@ -33,9 +34,12 @@
 	TTSFEATURE_VOLUME,
 	TTSMODEINFO,
 	VOICECHARSET,
+	waveOutGetNumDevs,
+	waveOutMessage,
+	DRV_QUERYFUNCTIONINSTANCEID,
+	DRV_QUERYFUNCTIONINSTANCEIDSIZE,
 )
 import config
-import nvwave
 import weakref
 
 from speech.commands import (
@@ -233,7 +237,7 @@ def _set_voice(self, val):
 			raise ValueError("no such mode: %s" % val)
 		self._currentMode = mode
 		self._ttsAudio = CoCreateInstance(CLSID_MMAudioDest, IAudioMultiMediaDevice)
-		self._ttsAudio.DeviceNumSet(nvwave.outputDeviceNameToID(config.conf["audio"]["outputDevice"], True))
+		self._ttsAudio.DeviceNumSet(_mmDeviceEndpointIdToWaveOutId(config.conf["audio"]["outputDevice"]))
 		self._ttsCentral = POINTER(ITTSCentralW)()
 		self._ttsEngines.Select(self._currentMode.gModeID, byref(self._ttsCentral), self._ttsAudio)
 		self._ttsAttrs = self._ttsCentral.QueryInterface(ITTSAttributes)
@@ -365,3 +369,37 @@ def _set_volume(self, val: int):
 		# using the low word for the left channel and the high word for the right channel.
 		val |= val << 16
 		self._ttsAttrs.VolumeSet(val)
+
+
+def _mmDeviceEndpointIdToWaveOutId(targetEndpointId: str) -> int:
+	if targetEndpointId != config.conf.getConfigValidation(("audio", "outputDevice")).default:
+		targetEndpointIdByteCount = (len(targetEndpointId) + 1) * sizeof(c_wchar)
+		currEndpointId = create_string_buffer(targetEndpointIdByteCount)
+		currEndpointIdByteCount = DWORD()
+		for devID in range(waveOutGetNumDevs()):
+			# Get the length of this device's endpoint ID string.
+			mmr = waveOutMessage(
+				HANDLE(devID),
+				DRV_QUERYFUNCTIONINSTANCEIDSIZE,
+				byref(currEndpointIdByteCount),
+				None,
+			)
+			if (mmr != MMSYSERR_NOERROR) or (currEndpointIdByteCount.value != targetEndpointIdByteCount):
+				# ID lengths don't match, so this device can't be a match.
+				continue
+			# Get the device's endpoint ID string.
+			mmr = waveOutMessage(
+				HANDLE(devID),
+				DRV_QUERYFUNCTIONINSTANCEID,
+				byref(currEndpointId),
+				currEndpointIdByteCount,
+			)
+			if mmr != MMSYSERR_NOERROR:
+				continue
+			# Decode the endpoint ID string to a python string, and strip the null terminator.
+			if (
+				currEndpointId.raw[: targetEndpointIdByteCount - sizeof(c_wchar)].decode("utf-16")
+				== targetEndpointId
+			):
+				return devID
+	return -1
@@ -4,6 +4,9 @@
 
 ### Important notes
 
+* Users are warned that support for Microsoft Speech API version 4 synthesizers is planned for removal in NVDA 2026.1.
+Any remaining users of SAPI4 speech synthesizers are encouraged to choose a more modern speech synthesizer.
+
 ### New Features
 
 * Support for math in PDFs has been added.
@@ -179,6 +182,7 @@ Use `gui.message.MessageDialog` instead. (#17582)
   * `NoConsoleOptionParser`, `stringToBool`, `stringToLang` in `__main__`; use the same symbols in `argsParsing` instead.
   * `__main__.parser`; use `argsParsing.getParser()` instead.
 * `bdDetect.DeviceType` is deprecated in favour of `bdDetect.ProtocolType` and `bdDetect.CommunicationType` to take into account the fact that both HID and Serial communication can take place over USB and Bluetooth. (#17537 , @LeonarddeR)
+* SAPI4, and all of its code, are deprecated, and planned for removal in 2026.1.
 
 ## 2024.4.2