Datasets:
audio audio | text string | source string | source_dataset string | source_index int64 | noise_file string | noise_start_sample int64 | target_snr_db float64 | achieved_snr_db float64 | peak_gain_db float64 | sampling_rate int64 | seed int64 |
|---|---|---|---|---|---|---|---|---|---|---|---|
Yeqqim d igellil kra yedder. | tatoeba | boffire/kabyle-piper-22khz | 0 | bruitages/campagne.wav | 692,830 | 10 | 10 | 0 | 22,050 | 20,260,924 | |
Bḍan. | tatoeba | boffire/kabyle-piper-22khz | 1 | bruitages/hall.wav | 1,171,707 | 10 | 10 | -0.08 | 22,050 | 20,260,924 |
Kabyle audio with noise — two-clip test
This test batch contains 10 WAV files generated from two streamed records of boffire/kabyle-piper-22khz using the noise-generation logic from boffire/kab-noise-asxerxec/ajouter_bruit.py.
Processing: mono, 22,050 Hz; two seconds of background-only audio before and after each voice; source voice normalized to -20 dBFS; primary noise randomly selected at -44 to -36 dBFS; secondary noise probability 30%; noise fades 400/600 ms; peak limited to -1 dBFS. The original Kabyle text, source, noise parameters, durations, and generation seed are in manifest.csv.
This is a smoke test, not an evaluation set. Verify the upstream audio and environmental-noise redistribution rights before producing a larger public batch.
- Downloads last month
- 76