Merge pull request #53 from rhasspy/revert-48-sentence-silence

Revert "add sentence silence"

Author: synesthesiam

tests/test_piper.py

@@ -19,7 +19,6 @@
 _DIR = Path(__file__).parent
 _LOCAL_DIR = _DIR.parent / "local"
 _TIMEOUT = 60
-_REBASELINE = False
 
 
 @pytest.mark.asyncio
@@ -34,8 +33,6 @@ async def test_piper() -> None:
         "en_US-ryan-low",
         "--data-dir",
         str(_LOCAL_DIR),
-        "--sentence-silence",
-        "0.5",
         stdin=PIPE,
         stdout=PIPE,
     )
@@ -61,7 +58,7 @@ async def test_piper() -> None:
 
     # Synthesize text
     await async_write_event(
-        Synthesize("This is a test. This is sentence two.", voice=SynthesizeVoice("en_US-ryan-low")).event(),
+        Synthesize("This is a test.", voice=SynthesizeVoice("en_US-ryan-low")).event(),
         proc.stdin,
     )
 
@@ -70,6 +67,13 @@ async def test_piper() -> None:
     assert AudioStart.is_type(event.type)
     audio_start = AudioStart.from_event(event)
 
+    with wave.open(str(_DIR / "this_is_a_test.wav"), "rb") as wav_file:
+        assert audio_start.rate == wav_file.getframerate()
+        assert audio_start.width == wav_file.getsampwidth()
+        assert audio_start.channels == wav_file.getnchannels()
+        expected_audio = wav_file.readframes(wav_file.getnframes())
+        expected_array = np.frombuffer(expected_audio, dtype=np.int16)
+
     actual_audio = bytes()
     while True:
         event = await asyncio.wait_for(async_read_event(proc.stdout), timeout=_TIMEOUT)
@@ -86,21 +90,6 @@ async def test_piper() -> None:
 
     actual_array = np.frombuffer(actual_audio, dtype=np.int16)
 
-    if _REBASELINE:
-        with wave.open(str(_DIR / "this_is_a_test.wav"), "w") as wav_file:
-            wav_file.setframerate(audio_start.rate)
-            wav_file.setsampwidth(audio_start.width)
-            wav_file.setnchannels(audio_start.channels)
-            wav_file.writeframes(actual_audio)
-        return
-    else:
-        with wave.open(str(_DIR / "this_is_a_test.wav"), "rb") as wav_file:
-            assert audio_start.rate == wav_file.getframerate()
-            assert audio_start.width == wav_file.getsampwidth()
-            assert audio_start.channels == wav_file.getnchannels()
-            expected_audio = wav_file.readframes(wav_file.getnframes())
-            expected_array = np.frombuffer(expected_audio, dtype=np.int16)
-
     # Less than 20% difference in length
     assert (
         abs(len(actual_array) - len(expected_array))

tests/this_is_a_test.wav

@@ -3,9 +3,9 @@
 import asyncio
 import json
 import logging
-import signal
 from functools import partial
 from pathlib import Path
+import signal
 from typing import Any, Dict, Set
 
 from wyoming.info import Attribution, Info, TtsProgram, TtsVoice, TtsVoiceSpeaker
@@ -54,11 +54,6 @@ async def main() -> None:
     parser.add_argument(
         "--noise-w-scale", "--noise-w", type=float, help="Phoneme width noise"
     )
-    parser.add_argument(
-        "--sentence-silence",
-        type=float,
-        help="Seconds of silence to add after each sentence",
-    )
     #
     parser.add_argument(
         "--auto-punctuation",

wyoming_piper/__main__.py

@@ -233,18 +233,6 @@ async def _handle_synthesize(
                 with wav_writer:
                     _VOICE.synthesize_wav(text, wav_writer, syn_config)
 
-                    if self.cli_args.sentence_silence is not None and not send_stop:
-                        # 16-bit samples for silence
-                        wav_writer.writeframes(
-                            bytes(
-                                int(
-                                    _VOICE.config.sample_rate
-                                    * self.cli_args.sentence_silence
-                                    * 2
-                                )
-                            )
-                        )
-
             output_file.seek(0)
 
             wav_file: wave.Wave_read = wave.open(output_file, "rb")