Update README.md
Browse files
README.md
CHANGED
@@ -33,7 +33,7 @@ The dataset contains metadata for approximately 36.7k hours of audio from 5k spe
|
|
33 |
|
34 |
The metadata contains estimated bandwidth, which can be used to infer the original sampling rate the audio was recorded at. The base dataset is filtered for a bandwidth appropriate for training speech models at 22 kHz. We also provide a precomputed subset with 31.7k hours appropraite for 44 kHz training. Users can modify the download script to use any sampling rate and bandwidth threshold which might be more appropriate for their work.
|
35 |
|
36 |
-
|
37 |
|
38 |
## Dataset Format
|
39 |
|
|
|
33 |
|
34 |
The metadata contains estimated bandwidth, which can be used to infer the original sampling rate the audio was recorded at. The base dataset is filtered for a bandwidth appropriate for training speech models at 22 kHz. We also provide a precomputed subset with 31.7k hours appropraite for 44 kHz training. Users can modify the download script to use any sampling rate and bandwidth threshold which might be more appropriate for their work.
|
35 |
|
36 |
+
LibriVox audiobooks are not redistributed on Hugging Face. All audio in the dataset can be downloaded from LibriVox, following the instructions below.
|
37 |
|
38 |
## Dataset Format
|
39 |
|