Merge pull request #977 from hanasay/main

Convert audio to mono while extract speech token
FunAudioLLM · Feb 16, 2025 · fd45708 · fd45708
2 parents 95e99e0 + 296ed4f
commit fd45708
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/tools/extract_speech_token.py b/tools/extract_speech_token.py
@@ -27,6 +27,9 @@ def single_job(utt):
     audio, sample_rate = torchaudio.load(utt2wav[utt], backend='soundfile')
     if sample_rate != 16000:
         audio = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)(audio)
+    # Convert audio to mono
+    if audio.shape[0] > 1:
+        audio = audio.mean(dim=0, keepdim=True)
     if audio.shape[1] / 16000 > 30:
         logging.warning('do not support extract speech token for audio longer than 30s')
         speech_token = []