Bug fix in MP3 length on TTSDataset

TTS/tts/datasets/dataset.py

@@ -13,6 +13,8 @@
 from TTS.utils.audio import AudioProcessor
 from TTS.utils.audio.numpy_transforms import compute_energy as calculate_energy
 
+from mutagen.mp3 import MP3
+
 # to prevent too many open files error as suggested here
 # https://github.com/pytorch/pytorch/issues/11201#issuecomment-421146936
 torch.multiprocessing.set_sharing_strategy("file_system")
@@ -42,6 +44,19 @@ def string2filename(string):
     return filename
 
 
+def get_audio_size(audiopath):
+    if audiopath[-4:] == ".mp3":
+        audio_info = MP3(audiopath).info
+        return int(audio_info.length * audio_info.sample_rate)
+    elif audiopath[-4:] == ".wav" or audiopath[-5:] == ".flac":
+        compress_factor = 8
+        bitrate = 16 # assuming 16bit audio
+        return int(os.path.getsize(audiopath) / bitrate * compress_factor)
+    else:
+        audio_format = audiopath.split(".")[-1]
+        raise RuntimeError(f"The audio format {audio_format} is not supported, please convert the audio files for mp3, flac or wav format!")
+
+
 class TTSDataset(Dataset):
     def __init__(
         self,
@@ -176,7 +191,7 @@ def lengths(self):
         lens = []
         for item in self.samples:
             _, wav_file, *_ = _parse_sample(item)
-            audio_len = os.path.getsize(wav_file) / 16 * 8  # assuming 16bit audio
+            audio_len = get_audio_size(wav_file)
             lens.append(audio_len)
         return lens
 
@@ -295,7 +310,7 @@ def load_data(self, idx):
     def _compute_lengths(samples):
         new_samples = []
         for item in samples:
-            audio_length = os.path.getsize(item["audio_file"]) / 16 * 8  # assuming 16bit audio
+            audio_length = get_audio_size(item["audio_file"])
             text_lenght = len(item["text"])
             item["audio_length"] = audio_length
             item["text_length"] = text_lenght

requirements.txt

@@ -17,6 +17,7 @@ pyyaml==6.*
 fsspec==2023.6.0 # <= 2023.9.1 makes aux tests fail
 aiohttp==3.8.*
 packaging==23.1
+mutagen==1.47.0
 # deps for examples
 flask==2.*
 # deps for inference