How to use Jaco-Assistant STT?


so I want to get DeepSpeech in German via Jaco-Assistant on my Rasperry Pi 4B, but I’m not quite sure how to download and use it. Should I just execute the Python-Program found at Or is there another way of getting it, e.g. via the terminal?
And when I have got the new models, how do I use them? Or in another way: how does the deepspeech --model deepspeech-0.9.3-models.tflite --scorer deepspeech-0.9.3-models.scorer --audio audio/2830-3980-0043.wav command change?
Sorry, if there already is an answer somewhere, I couldn’t find it.

Thank you already.

Jaco is great, but not my strong suit. @dan.bmh should be able to help :slight_smile:

If you’re just interested in the STT features you can download the models (.tflite + .scorer) directly from the DS-Polyglot repo:

Jaco also handles NLU interpretation and automatic scorer building from the skill data, but it seems that’s more than you currently need.

1 Like

So I’m just going to get the model via curl -LO and the scorer via curl -LO
But when I’m trying to use it with deepspeech --model output_graph_de.tflite --scorer kenlm_de.scorer --audio audio/2830-3980-0043.wav I get an Error ERROR: Could not open 'output_graph_de.tflite'.
How can I solve this? / What did I do wrong?

Are you sure the files are not broken?
In the script you linked above I did need a special approach to download the files from google drive.

With the curl commands above I only get some html files.

This is possible. How did you do it in that “special way”?

Please use a search engine (e.g. Google) to find out how to download gdrive files directly. This has nothing to do with Jaco nor DeepSpeech and I remember that you have to use some trick, but don’t remember it.

Please look at the script you linked in your first post.
With some edits you should be able to use parts of it without installing the rest of Jaco.