DeepSpeech with Common Voice Training Data

DeepSpeech
#1

Does the DeepSpeech Model downloaded from the below link is trained with Common Voice Training Data or We need to train and extract the model separately

curl -LO https://github.com/mozilla/DeepSpeech/releases/download/v0.5.1/deepspeech-0.5.1-models.tar.gz

#2

The 0.5.1 model wasn’t trained on Common Voice data.

#3

Thanks for reply @dabinat. Whether we can get trained model on Common Voice Data from Online

#4

What is the question here ?

#5

Whether we can get readymade model which is already trained on Common Voice Data

#6

You have to do it, like we did for Italian https://github.com/MozillaItalia/DeepSpeech-Italian-Model/ or for french https://github.com/Common-Voice/commonvoice-fr/

#7

Hi, i’d like to train a ASR and a TTS model for Italian on Common Voice Training Data. I have few questions:

  1. Is there already some model available readyToUse? I look at the github repo you link but there’s just the link to DeepSpeech “CodeTrainer” and a tool to manipulate datas. That is stuff just to train own’s model, right?
  2. is this a good step to start with to understand your architecture? https://arxiv.org/abs/1412.5567
  3. Sorry but there’s really a lot of infos on this site forum that i got lost. To train my own model have i to follow this guide TUTORIAL : How I trained a specific french model to control my robot ?
  4. Does DeepSpeech works well only for ASR or also for TTS ? In case one can you suggest also an algorithm or paper or architecture for TTS?
  5. Is some kind of preprocessing required?

Thank you in advance if you would like to help me.

#8

Hi Christian we have the italian category for that https://discourse.mozilla.org/c/voice/it