microsoft/VibeVoice · error · ValueError

Empty audio list provided

Error message

Empty audio list provided

What it means

When __call__ receives a Python list, it immediately requires it to be non-empty. An empty list cannot be dispatched (the code inspects audio[0] to distinguish paths from arrays), so it fails fast instead of indexing out of bounds or silently returning an empty batch.

Source

Thrown at vibevoice/processor/vibevoice_tokenizer_processor.py:159

        """
        if audio is None:
            raise ValueError("Audio input is required")
        
        # Validate sampling rate
        if sampling_rate is not None and sampling_rate != self.sampling_rate:
            logger.warning(
                f"Input sampling rate ({sampling_rate}) differs from expected "
                f"sampling rate ({self.sampling_rate}). Please resample your audio."
            )
        
        # Handle different input types
        if isinstance(audio, str):
            # Single audio file path
            audio = self._load_audio_from_path(audio)
            is_batched = False
        elif isinstance(audio, list):
            if len(audio) == 0:
                raise ValueError("Empty audio list provided")
            
            # Check if it's a list of file paths
            if all(isinstance(item, str) for item in audio):
                # Batch of audio file paths
                audio = [self._load_audio_from_path(path) for path in audio]
                is_batched = True
            else:
                # Check if it's batched audio arrays
                is_batched = isinstance(audio[0], (np.ndarray, list))
        else:
            # Single audio array or list
            is_batched = False
        
        # Process audio
        if is_batched:
            processed_audio = [self._process_single_audio(a) for a in audio]
        else:
            processed_audio = [self._process_single_audio(audio)]

View on GitHub (pinned to 94da20d98b)

Solutions

1. Skip the call when the batch is empty: if not audio_list: continue.
2. Fix upstream filters/loaders that can produce zero items so empty batches are surfaced as their own error.
3. For single-item cases, pass the item directly rather than a one-or-zero-element list.

Example fix

# before
enc = processor(audio=loaded)  # loaded == [] after filter

# after
if not loaded:
    continue
enc = processor(audio=loaded)

if isinstance(audio, list) and len(audio) == 0:
    raise ValueError('empty audio batch — upstream filter removed all items')
enc = tokenizer_processor(audio=audio)

def is_nonempty_audio_batch(audio) -> bool:
    return not isinstance(audio, list) or len(audio) > 0

try:
    enc = tokenizer_processor(audio=batch)
except ValueError as e:
    if 'Empty audio list' in str(e):
        logger.warning('skipping empty batch')
        return None
    raise

When it happens

Trigger: Calling processor(audio=[]) — e.g. a manifest-driven pipeline filtered out all files for a shard, or a batch loader dropped failed items and returned [].

Common situations: Batch inference loops over dataset shards where one shard ends up empty after filtering; error-handling code that collects successfully loaded audio and sometimes collects nothing; placeholder code awaiting real data.

Related errors

• Sample index {sample_idx} exceeds batch size {self.batch_siz
• Audio input is required for ASR processing
• Audio input is required
• Unsupported file format: {file_ext}. Supported formats: .wav
• load_audio_bytes_use_ffmpeg requires resample=True

AI-assisted analysis of microsoft/VibeVoice@94da20d98b (2026-08-15). Data as JSON: /api/errors/9666bc7d501ae7dd. Report an issue: GitHub.