A list of common publically (and privately) available audio data that you can download for ASR or other speech activities. All your WERs are belong to us. Inspired by wer are we who stole someone elses joke.
Source | Name & Direct Link | Type | Size(Hours) |
---|---|---|---|
OpenSLR | LibriSpeech - Train:100 360 500 Test:Clean Other Dev:Clean Other |
Read | 960 |
OpenSLR | TED-LIUM Release 2 | Read | 118 |
OpenSLR | TED-LIUM Release 3 | Read | 452 |
Voxforge | Voxforge English | Read | 130 |
Mozilla | Common Voice v1 | Read | 500 |
Mozilla | Common Voice en_1087h_2019-06-12 | Read | 1087 |
Tatoeba | Tatoeba Audio Eng | Read | ~200 |
Valentini | Noisy Speech Database All Files, DOI | Read | TBC |
Source | Name | Type | Size(Hours) | Code |
---|---|---|---|---|
LDC | Fisher | Conversational | 2000 | Speech LDC2004S13 LDC2005S13 Transcripts LDC2004T19 LDC2005T19 |
LDC | Switchboard Hub 500 | Conversational | 240 | LDC2002S09 |
LDC | Switchboard Release 2 | Conversational | 300 | LDC97S62 |
LDC | TIMIT | Read | 5 | LDC93S1 |
LDC | Wall Street Journal (WSJ) | Read | 80 | LDC93S6A or LDC93S6B |
Source | Name & Direct Link | Type | Size(Hours) |
---|---|---|---|
Edinburgh CSTR | CSTR VCTK Corpus | Read | 44 |
LJ Speech | LJ Speech | Read | 24 |