search

alumae / kaldi-gstreamer-server

Real-time full-duplex speech recognition server, based on the Kaldi toolkit and the GStreamer framwork.
BSD 2-Clause "Simplified" License
1.07k stars 339 forks source link

Any updates on the English model? #99

Open mertyildiran opened 7 years ago

mertyildiran commented 7 years ago

Is there any update coming up on the English model? (voxforge/tri2b_mmi_b0.05)

According to history, the model is 4 years old. I suppose there must be some improvements, at least on VoxForge end.

Also from where did you exactly get this model? The download link in here seems like broken. Which one of these models is the successor of the current model?

Finally, how can we train our own model to use it with kaldi-gstreamer-server? Is there a guide/tutorial for that?

mertyildiran commented 7 years ago

Also I forgot to mention; with Git LFS now you can have files up to 2 GB in size. (About Git Large File Storage)

So you don't have to actually ensure "100 MB GitHub file size limit" criteria if you are not planning to track the history of the files and you don't have to track the history of the model because the files are binary. (you can exclude words.txt which is a text file)

Another alternative to that is; distributing the models via releases section.

We don't limit the total size of your binary release files, nor the bandwidth used to deliver them. However, each individual file must be under 2 GB in size. - Distributing large binaries