Hey everyone! π
I'm currently building a little learning app for my son, and Iβve run into a snag with text-to-speech on Windows.
The app works great overall, but while it picks up the correct voice on macOS (say), on Windows it seems to ignore the HD Narrator/Natural voice we installed and defaults to the generic system voice instead.
Here is the code Iβm using for the speech synthesis:
@staticmethod
def speak(text: str) -> None:
"""Synthesize and play text-to-speech using OS native APIs.
Runs on background thread to avoid blocking UI. Supports macOS (say) and Windows (PowerShell).
Args:
text: Text string to speak aloud.
"""
def _run_speech() -> None:
"""Execute speech synthesis on background thread."""
system_name = platform.system()
try:
# macOS uses native 'say' command.
if system_name == "Darwin":
subprocess.run(["say", text], check=False)
# Windows uses System.Speech which respects the user's selected narrator voice.
elif system_name == "Windows":
safe_text = text.replace("'", "''") # Escape single quotes for PowerShell.
cmd = (
"Add-Type -AssemblyName System.Speech; "
"$synth = New-Object System.Speech.Synthesis.SpeechSynthesizer; "
f"$synth.Speak('{safe_text}');"
)
subprocess.run(
["powershell", "-NoProfile", "-Command", cmd],
check=False,
creationflags=0x08000000, # CREATE_NO_WINDOW β keeps it silent in background.
)
else:
print(f"Speech not supported on platform: {system_name}")
except (OSError, subprocess.SubprocessError) as e:
print(f"Speech error: {e}")
# Start speech in daemon thread so UI remains responsive.
threading.Thread(target=_run_speech, daemon=True).start()
Has anyone run into an issue with System.Speech.Synthesis.SpeechSynthesizer in PowerShell failing to pick up newly installed Windows HD/Natural voices? Do I need to explicitly select the voice by name in the PowerShell script rather than relying on the default?
Any tips or pointers would be massively appreciated! π
Top comments (3)
System.Speech only sees classic SAPI5 voices, so the HD Narrator ones never show up there and Windows drops back to the desktop default. That's why say on the Mac follows the voice you picked and this PowerShell path doesn't. GetInstalledVoices() on that synthesizer will list what this code can actually use. The Natural voices need Windows.Media.SpeechSynthesis.SpeechSynthesizer, not System.Speech.
Ah, thank you for explaining that! I appreciate the help and the information. I'll have a go at trying this out when I get a chance. Thanks again!
Some comments may only be visible to logged-in visitors. Sign in to view all comments.