Pretrained models for speech recognition

Here you will find models for speech recognition of languages in Common Voice. The models have been trained with Coquí STT.

AcousticAcoustic + LMTraining
LanguageCodeCERWERCERWERLossAudio (h:m:s)Clips
Gaeilgega-IE40.572186.882942.116470.727455.1607320:31:24542
Suomifi30.693696.647427.921360.539553.3717310:32:29456
ଓଡ଼ିଆor34.998895.000036.050774.583361.0971070:32:56389
Hakha Chincnh26.479967.364424.646053.276333.3964690:38:14807
Rumantschrm-vallader26.217084.007821.587954.284157.9865420:58:58558
Чӑвашлаcv33.727095.374233.099564.979258.8938671:04:44932
Lietuvių kalbalt31.049294.638529.455067.218161.8997841:10:17928
Hornjoserbšćinahsb32.432092.324432.227166.575977.5217061:26:58799
Саха тылаsah36.327794.503739.570371.997284.3127371:27:01918
Lugandalg30.477693.129328.401463.206358.3235861:36:091251
ქართული ენაka31.126295.751828.087359.831866.7799151:37:211055
Türkçetr30.835889.263129.623857.187444.8567541:52:311739
Brezhonegbr37.711989.118838.133768.366937.9944112:06:292781
Romontschrm-sursilv23.881979.566518.928348.066545.2393042:07:241381
Bahasa indonesiaid25.793480.717316.061332.667834.5370182:09:492131
Slovenščinasl26.786682.362318.163040.329032.8176542:15:112038
Latviešu valodalv28.306982.814816.422832.955229.1760222:17:052553
தமிழ்ta46.5789 99.929949.3607100.000061.5451552:21:152010
Maltimt27.917586.398921.951546.888551.0234532:31:232037
Кыргызчаky30.549487.065226.325852.190153.0203782:34:341955
Ελληνικάel31.196680.212324.353648.837243.0922742:45:042314
Монгол хэлmn38.606490.802438.155669.001890.0214463:02:382168
ภาษาไทยth35.9950100.000051.5506100.000042.4703183:24:572915
Româneştero27.999082.119218.546136.339547.8899613:37:293369
ދިވެހިdv27.440088.368122.230866.492062.7815093:56:122632
Magyar nyelvhu31.003185.868922.277144.270947.8305364:17:043339
Eestiet24.988285.526519.616346.049473.7160875:00:162760
Fryskfy-NL26.491374.046819.767341.204546.3959735:26:023923
Portuguêspt26.685173.153820.098439.713843.5927937:40:456319
Euskaraeu15.646768.69106.987120.640332.83897810:51:347505
Татарчаtt31.682885.812426.384553.222644.46754811:49:1511181

What is Coquí STT?

Coquí STT is the only privacy-respecting on-device end-to-end speech recognition system in existence. Most speech recognition systems either require you to broadcast your personal information over the internet or require expertise in machine learning or computational linguistics to set up. Coquí STT works on your device, respects your privacy, is easy to set up, easy to adapt and easy to deploy.