HiFiTTS-2 Bandwidth Estimation

This config contains the bandwidth estimation code used for HiFiTTS and HiFiTTS-2. This config can be used to estimate bandwidth for any dataset. For HiFiTTS-2 bandwidth was estimated using the first 30 seconds of every audiobook chapter, but the estimate is still reasonably accurate if run over a shorter duration or with individual utterances.

Required arguments.

• workspace_dir: The workspace folder where all audio files and manifests are stored.

• audio_dir: Folder in workspace containing audio files to estimate bandwidth of.

• input_manifest_filename: Manifest file in workspace containing relative paths to audio.

Output format.

This config outputs a single manifest with the following field(s):

• bandwidth (int): Estimated bandwidth of the audio file.

Config link: dataset_configs/english/hifitts2/config_bandwidth.yaml