chore: add utils/preview_effects.py for A/B comparison

utils/preview_effects.py

@@ -0,0 +1,55 @@
+import os
+import sys
+
+import numpy as np
+import soundfile as sf
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+import config
+import effects
+import tts
+
+
+PHRASE = "Приветствую, сэр. Все системы в норме."
+OUT_RAW = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
+                      'sound', '_preview_silero_raw.wav')
+OUT_FX = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))),
+                     'sound', '_preview_silero_fx.wav')
+
+
+def _render_raw(text: str) -> np.ndarray:
+    audio = tts.model.apply_tts(text=text + "..",
+                                speaker=tts.speaker,
+                                sample_rate=tts.sample_rate,
+                                put_accent=tts.put_accent,
+                                put_yo=tts.put_yo)
+    return np.asarray(audio, dtype=np.float32)
+
+
+def _render_fx(raw: np.ndarray) -> np.ndarray:
+    return effects.process(
+        raw, tts.sample_rate,
+        bandpass=(config.TTS_BANDPASS_LOW_HZ, config.TTS_BANDPASS_HIGH_HZ),
+        reverb=(config.TTS_REVERB_WET, config.TTS_REVERB_DECAY_MS),
+        pitch_semitones=config.TTS_PITCH_SEMITONES,
+    )
+
+
+def main():
+    print(f"Synthesizing: {PHRASE!r}")
+    raw = _render_raw(PHRASE)
+    fx = _render_fx(raw)
+
+    sf.write(OUT_RAW, raw, tts.sample_rate, subtype='PCM_16')
+    sf.write(OUT_FX, fx, tts.sample_rate, subtype='PCM_16')
+
+    dur_raw = len(raw) / tts.sample_rate
+    dur_fx = len(fx) / tts.sample_rate
+    print(f"raw: {OUT_RAW}  ({len(raw)} samples, {dur_raw:.2f}s)")
+    print(f"fx : {OUT_FX}  ({len(fx)} samples, {dur_fx:.2f}s)")
+    print("Open both in a media player and A/B compare.")
+
+
+if __name__ == '__main__':
+    main()