---
configs:
- config_name: default
  data_files:
  - split: train
    path: train/*.parquet
---

# AVSpeech Video + Audio

This dataset stores AVSpeech media as separate columns:
- `video`: video-only stream (no audio), stream-copied without re-encoding.
- `audio`: audio-only stream, stream-copied without re-encoding.