AVSynthesiser for Mac Engine

[willwade]

So this looks ok to me. I just need some brains to figure out how we

• Add to test suite
• Checking docs
• Double checking code and making sure tests work

#347 #336

I think this is passing all the tests. Its far simpler than nsss and we can do thngs like ssml #121 #287

[cclauss]

Edit

pyttsx3/pyttsx3/driver.py

Line 41 in da247c1

'darwin': 'nsss',

1. Change nsss to avsynth and save the change
2. python -m pip install --editable .
3. python

>>> import pyttsx3
>>> engine = pyttsx3.init()

[willwade]

Yeah. Slightly nervous of this. I need to make sure all tests are good and stuff like voices are the same. Don't want to break legacy code

[cclauss]

Local tests work for me.

[cclauss]

>>> engine=pyttsx3.init('nsss')
>>> engine.say("This is a test")
>>> engine.runAndWait()  # This works and the REPL does not quit.
>>> engine=pyttsx3.init('avsynth')
>>> engine.say("This is a test")
>>> engine.runAndWait()  # This works but the REPL quits which is a BUG.

[cclauss]

Will avsynth work on iOS?

[willwade]

Will avsynth work on iOS?

well "technically" it should.. but no idea how you get all of python etc working on ios like this..

[willwade]

failing on external event loop test. working on a fix..

[cclauss]

This test can be added near the top of tests/test_pyttsx3.py before the other tests mess with the engine. The test passes for me but unfortunately, I hear no utterences.

@pytest.mark.skipif(sys.platform not in ("darwin", "ios"), reason="Testing only on macOS and iOS")
def test_apple_avsynth_voices(engine):
    import platform

    macos_version, _, macos_hardware = platform.mac_ver()
    print(f"{sys.platform = }, {macos_version = } on {macos_hardware = }")
    print(list(pyttsx3._activeEngines))
    print(engine)
    # assert str(engine) == "avsynth", "Expected engine name to be avsynth on macOS and iOS"
    voice = engine.getProperty("voice")
    # On macOS v14.x, the default nsss voice is com.apple.voice.compact.en-US.Samantha.
    # ON macOS v15.x, the default nsss voice is "".
    # ON macOS v15.x, the default avsynth voice is None.

    assert voice in (None, "", "com.apple.voice.compact.en-US.Samantha"), (
        f"Expected default voice {voice} to be com.apple.voice.compact.en-US.Samantha"
    )
    voices = engine.getProperty("voices")
    # On macOS v14.x, nsss has 143 voices.
    # On macOS v15.x, nsss has 176 voices
    print(f"On macOS v{macos_version}, {engine} has {len(voices) = } voices.")
    assert len(voices) in (176, 143), "Expected 176 or 143 voices on macOS and iOS"
    # print("\n".join(voice.id for voice in voices))
    en_us_voices = [voice for voice in voices if voice.id.startswith("com.apple.eloquence.en-US.")]
    assert len(en_us_voices) == 8, "Expected 8 com.apple.eloquence.en-US voices on macOS and iOS"
    names = []
    for _voice in en_us_voices:
        engine.setProperty("voice", _voice.id)
        name = _voice.id.split(".")[-1]
        names.append(name)
        engine.say(f"{name} says hello.")
    name_str = ", ".join(names)
    assert name_str == "Eddy, Flo, Grandma, Grandpa, Reed, Rocko, Sandy, Shelley"
    print(f"({name_str})", end=" ", flush=True)
    engine.runAndWait()
    engine.setProperty("voice", voice)  # Reset voice to original value

[willwade]

Yeah. This is definitely broken right now. This and cleaning up the espeak stuff is on my priority list

[cclauss]

Suggested change

	print(
	f"Finished utterance: {utterance.speechString()}"
	) # Debugging: Track each completed utterance
	# Debugging: Track each completed utterance
	print(f"Finished utterance: {utterance.speechString()}")

[cclauss]

Run pre-commit run --all-files to fix the pre-commit error.

[cclauss]

Suggested change

	assert voice in (
	None,
	"",
	"com.apple.voice.compact.en-US.Samantha",
	), f"Expected default voice {voice} to be com.apple.voice.compact.en-US.Samantha"
	if voice:
	assert (
	voice == "com.apple.voice.compact.en-US.Samantha",
	), f"Expected default voice {voice} to be com.apple.voice.compact.en-US.Samantha"

[willwade]

Where we are up to:

• say and runandwait does nothing
  but if you say and say twice - it will work

I have something wrong with queuing

eg

# Initialize the pyttsx3 engine
engine = pyttsx3.init("avspeech")

def demo_play_simple(): 
    print("\nRunning demo_play_simple...")
    engine.say("Playing a simple message")
    engine.runAndWait()
    
    # Demo for testing multiple `say` calls followed by `runAndWait`
def demo_multiple_say_calls():
    print("\nRunning demo_multiple_say_calls...")
    engine.say("The first sentence.")

    print("Calling say after the first runAndWait()...")
    engine.say("The second sentence follows immediately.")

    print("Calling say after the second runAndWait()...")
    engine.say("Finally, the third sentence is spoken.")
    engine.runAndWait()  # Should speak "Finally, the third sentence is spoken."

demo_play_simple()
# Run the multiple `say` calls demo
demo_multiple_say_calls()

You will hear some of the sentences but not the final one. Its all out of sync