Models
This page contains Kaldi models available for download as .tar.gz archives. They may be downloaded and used for any purpose. Older models can be found on the downloads page. If you have models you would like to share on this page please contact us.
| Resource | Name | Category | Summary |
|---|---|---|---|
| M1 | ASpIRE Chain Model | ASR | A chain model trained on multi-condition Fisher English |
| M2 | CVTE Mandarin Model | ASR | Mandarin TDNN chain models trained on commercial data |
| M3 | SRE16 Xvector Model | SID | An xvector DNN trained on augmented LDC corpora |
| M4 | ASpIRE SAD Model | SAD | A TDNN used for speech activity detection |
| M5 | Tedlium Language Models | LM | LMs trained on Cantab-Tedlium text data and tedlium acoustic training data |
| M6 | Callhome Diarization Xvector Model | DIAR | An xvector DNN trained on augmented LDC corpora |
| M7 | VoxCeleb Models | SID | Pretrained wideband x-vector and i-vector systems |
| M8 | SITW Models | SID | Systems trained on VoxCeleb 1 and 2 for Speakers in the Wild |
| M9 | MGB-2 Arabic | ASR | A chain model developed for the MGB-2 challenge |
| M10 | DataTang Mandarin ASR System | ASR | A Mandarin ASR system developed by DataTang (Beijing) Co.Ltd. |
| M11 | Multi_CN ASR Model | ASR | A Mandarin ASR model, trained on free data |
| M12 | Chime 6 Models | SAD,DIAR,ASR | Pretrained SAD, diarization, and ASR baseline systems for the Chime 6 challenge |
| M13 | Librispeech ASR model | ASR | An English ASR model, trained on Librispeech (960h) |
| M14 | GigaSpeech ASR model | ASR | An English ASR model, trained on GigaSpeech |
