Datasets:

instinct-org
/

audio_youtube_chunked_qwen3_tts_12hz_tokenized

Languages:

Size:

Tags:

License:

Dataset card Data Studio Files Files and versions

You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

audio_youtube_chunked Qwen3-TTS 12Hz tokenized

This dataset contains Qwen3-TTS 12Hz tokenized JSONL shards for audio_youtube_chunked from the OmniVoice NFA-aligned all-dataset preparation run.

It is separate from the existing non-Qwen audio_youtube_chunked_tokenized artifacts. This repo contains tokenized JSONL records with audio_codes and metadata; raw audio files are not included.

Tokenizer: Qwen/Qwen3-TTS-Tokenizer-12Hz
Rows: 559475
Shards: 18
Published: 2026-05-28T20:57:59Z
Source run root: /root/runs/qwen3_tts_all15_nfa_tokenization_v1

Files are under data/*.jsonl; manifest.jsonl and dataset_summary.json provide row counts, sizes, and SHA-256 checksums.

Downloads last month: 25

Total file size: