text
stringlengths
2
5
6454
723
13031
1298
8141
3215
13022
2500
1915
2515
2501
3238
2518
12990
11048
107
6966
13021
12913
2508
12929
1918
2489
12906
10275
2093
7239
12947
7236
12972
12898
2525
8142
1618
248
1614
12928
12918
12963
2526
11049
12980
4018
128
1301
8143
12927
7232
12950
12983
1304
12926
8160
13009
13243
12993
688
249
12904
1594
12986
13003
13019
8151
12944
1324
12945
1296
2493
2495
1905
8152
10271
12949
1607
250
3204
1593
10279
3221
12924
12998
10276
252
8163
1649
12897
1603
12933
12905
13005
8158
2522
10270
13034
12955
2494
12912
3241
1916

The WhisperSpeech Dataset

This dataset contains data to train SPEAR TTS-like text-to-speech models that utilized semantic tokens derived from the OpenAI Whisper speech recognition model.

We currently provide semantic and acoustic tokens for the LibriLight and LibriTTS datasets (English only).

Acoustic tokens:

  • 24kHz EnCodec 6kbps (8 quantizers)

Semantic tokens:

  • Whisper tiny VQ bottleneck trained on a subset of LibriLight

Available LibriLight subsets:

  • small/medium/large (following the original dataset division but with large excluding the speaker 6454)
  • a separate ≈1300hr single-speaker subset based on the 6454 speaker from the large subset for training single-speaker TTS models

We plan to add more acoustic tokens from other codecs in the future.

Downloads last month
4
Edit dataset card

Space using collabora/whisperspeech 1