Only English Language Model

For my current use-case, language models other than English are not needed. Can I have a slim English language model? I read somewhere that DeepSpeech has a very lightweight English model that transcribes ‘faster than real-time’. Please point me to that direction. The current model is too large for my application.

Released language model is only for English, so you need to explain.

That’s very vague. Can you explain more?

Is it the acoustic model? the language model? what is your application?

In this comment, combined size of AM and LM is less than 100MB. Please suggest where I can get that LM model.

Ah, you are talking about the tflite model that can be used on mobile devices. You can get that on the release page and please search the forum a bit before asking questions :slight_smile:

As far as I know there is no smaller scorer in the releases, create your own with even more pruning and higher a,q bits.

thanks for clarifying. I was looking for the smaller scorer. OK

That was just an experiment ran by @reuben. You still have not clarified your needs and application so we can help you …