Datasets:

igorcouto
/

yt_health_voxtral_aug_both

Modalities:

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.

YAML Metadata Warning: empty or missing yaml metadata in repo card (https://huggingface.co/docs/hub/datasets-cards)

Model: mistralai/Voxtral-Mini-3B-2507 (128 mel, hop 160, n_fft 400, 16 kHz, padded to 30 s / 3000 frames)
Source: igorcouto/yt_health_audio_5_22s
Augmentations:
- aggressive: ON (1 variant/sample)
- far_echo : ON (1 variant/sample) ← far mic + echo + voice distortion
Original audio stored: YES
Columns: id, parent_id, is_aug, augment_preset, augment_seed, audio, text, log_mel[128,3000], language, start_ms, end_ms, segment_duration_s, avg_conf, low_conf_frac, channel_id, video_id, video_title, youtube_title, youtube_url