WaveWash: Clean, Repair, and Dedupe Audio Datasets Before ML Training

Описание и применение

WaveWash автоматически очищает, исправляет и удаляет дубликаты аудиофайлов, создавая структурированный набор данных для обучения моделей распознавания речи, синтеза речи и идентификации говорящих.

Задача

Training speech models often fails due to dirty audio data such as clipping, DC offset, background noise, low SNR, and duplicate samples that inflate validation metrics.

Результат

WaveWash automates audio cleaning, repair, and deduplication, producing a structured dataset with clean, error, and duplicate subsets and detailed logs, enabling developers to skip manual listening and scripting.

Рабочий процесс

• Install WaveWash via pip
• Launch the local web UI with wavewash app or run from terminal
• Select input directory and output directory
• Choose purpose preset (e.g., asr)
• Run the process
• Review segregated folders and logs

Доказательства

Training speech models (Whisper, TTS, voice cloning) often breaks down because of dirty audio data: clipping, DC offset, background noise, low SNR, and duplicate samples quietly inflating validation metrics.

Источник

I built WaveWash to automate audio quality control and data prep so developers don't have to write one-off scripts or manually listen to thousands of audio files.

Источник

Zero-Config Local Web UI: Ships directly inside the Python package — no Node.js build required.

Источник