Deepspeech multilanguage server

[JRMeyer]

>>> bruno-fischer
[December 10, 2020, 7:44am]

Good morning,

I have a basic question about using DeepSpeech.

I have the following project. I am setting up an Ubuntu 20.04 LTS server
(headless) and want to install DeepSpeech there. slash
The task of the server should be to send audio files (vmtl ogg, opus)
from a program via network to the server and let it process them. I do
not know which language the audio files contain. As return I want to
have the recognized text. I want to continue working with this in the
program.

How can I implement this? The program I use to deliver the data has a
Python interface.

Do I have to download different language models, if so where can I find
them?

Greetings from Berlin slash
Bruno

[This is an archived TTS discussion thread from discourse.mozilla.org/t/deepspeech-multilanguage-server]

[JRMeyer]

>>> othiele
[December 10, 2020, 9:10am]

> How can I implement this?

DeepSpeech does not recognize the language (de/en/...) for you, maybe
have several models transcribe your audio and see what gets good
confidence values?

> Do I have to download different language models, if so where can I
> find them?

Please search before you ask, have a look at this discourse and the
release and you'll quickly find all available models.

DeepSpeech has Python bindings, so you can easily load several models.
Look at the docs and the
examples.

[Archived Post]

[JRMeyer]

>>> bruno-fischer
[December 10, 2020, 1:47pm]

! slash
Thanks for your input.

I will look the links.

[Archived Post]