Skip to content

npentrel/babelfish

Repository files navigation

babelfish

Using speech to text in combination with text translation and text to speech, we built a tool not unlike Google's new babel platform. Call someone and they will hear what you say in their language 😮.

Both parties call the service's Nexmo number which will connect them to one another. However, both parties do not hear each other directly. They hear a translated version of one another thus enabling them both to speak and hear the other side in their preferred language. The way that works is that the audio stream of the speech on one caller's side will be transcribed and translated by the Microsoft Translator Speech API and the resulting text will be spoken to the other person. A German could thus hear a British person speaking German and respond in German. The response would be heard by the British person in English.

Stack

Python, Microsoft Translator Speech API, Nexmo Voice API, Ngrok.

Setup:

virtualenv venv
source venv/bin/activate
vim requirements.txt
pip install -r requirements.txt

In another terminal window run:

ngrok http 5000

Creating a Nexmo account

  • Create a Nexmo account
  • Install Nexmo CLI
  • Create an application
      nexmo app:create "Babelfish" http://your_url.ngrok.io/ncco http://your_url.ngrok.io/event --keyfile private.key
  • Purchase a number
      nexmo number:buy --country_code GB --confirm
  • Link your number and application
      nexmo link:app <number> <application_id>

Creating a Microsoft cognitive services account

  • Create a Microsoft Azure account
  • Search for Cognitive Services in the search bar
  • Click add and create a new Translator Speech API
  • Once created, click on the new translator service, then click on Keys under the Resource Management header
  • Copy KEY 1 for use in your application

Requirements

Configuring the application

Create config.py with the following contents:

HOSTNAME="your_host.ngrok.io" # Do not add HTTP
CALLER="<your_number>"
LANGUAGE1 = 'de-DE' # the caller's language
VOICE1 = 'Marlene' # a Nexmo voice for the caller's language

LANGUAGE2 = 'en-US' # the other person's language
VOICE2 = 'Kimberly' # a Nexmo voice for the other person's language

Create secrets.py with the following contents:

NEXMO_API_KEY = "<your-api-key>"
NEXMO_API_SECRET = "<your-api-secret>"
NEXMO_NUMBER = "+447512345678"
NEXMO_APPLICATION_ID = "<nexmo-application-id>"
NEXMO_PRIVATE_KEY = '''-----BEGIN PRIVATE KEY-----
<your-private-key>
-----END PRIVATE KEY-----'''

# You will have to sign up for a free Microsoft account to use the Microsoft Translator Speech API: http://docs.microsofttranslator.com/speech-translate.html
MICROSOFT_TRANSLATION_SPEECH_CLIENT_SECRET = "<your-api-key>"

Running the application

Run python main.py then have two (or more) people call your Nexmo number. When one person speaks, the rest will hear in another language. To change the languages used, edit config.py.

About

Call speech translation service

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages