Skip to content

Dataset · k9cli

Data shard and pipeline for audio extraction from video for ASR and TTS

k9cli/video-vec2wav2-tokenizer-2

Continuation of the project slicing videos into audio clips with Whisper transcriptions for speech tasks.

What's inside

The repository represents the second part (shard) of the video-to-AI-dataset tokenizer project and includes a ready-to-use Python pipeline. The tool turns a folder of video files into 16 kHz PCM mono audio, performs transcription via Whisper, and slices clips. The output generates metadata.csv, dataset.jsonl, tts_metadata.csv, and report.json files.

What you can do with it

  • Training and fine-tuning automatic speech recognition (ASR) models
  • Preparing datasets for text-to-speech (TTS) synthesis from raw video archives
  • Automating the collection and annotation of speech corpora via the video2dataset CLI utility
Downloads this month
448.7K
Likes
31

How to load it

from datasets import load_dataset

ds = load_dataset("k9cli/video-vec2wav2-tokenizer-2")