default noise_scale=0.0 (stable long-sequence decoding)

python/infer_board.py

@@ -221,7 +221,7 @@ def compress_pauses(wav, sr=24000, min_pause_ms=90, target_ms=60):
     return out
 
 
-def synthesize(acoustic, vocoder, text, noise_scale=0.3, seed=0):
+def synthesize(acoustic, vocoder, text, noise_scale=0.0, seed=0):
     sents = split_sentences(text)
     pieces = []
     for s in sents:
@@ -247,7 +247,7 @@ def main():
     parser.add_argument("--acoustic", default="export/acoustic_female.onnx")
     parser.add_argument("--vocoder", default="export/axmodel/bigvgan_base.axmodel")
     parser.add_argument("--output", default="board_out.wav")
-    parser.add_argument("--noise_scale", type=float, default=0.3)
+    parser.add_argument("--noise_scale", type=float, default=0.0)
     args = parser.parse_args()
 
     import soundfile as sf