Skip to content

Batch translate audio languages

This document shows how to use the Speech-to-Speech Translation Batch solution, designed to translate speech audio from one language to another.

Summary

The Speech-to-Speech Translation solution leverages Google Cloud services, including Speech-to-Text, Translation API, Text-to-Speech, as well as Cloud Storage to perform the following chained steps:

  1. Speech-to-Text: The source audio file is transcribed using the Speech-to-Text API, converting it to text.
  2. Translation: The transcribed text is translated to the target language using the Translation API.
  3. Text-to-Speech: The translated text is synthesized into speech using the Text-to-Speech API, generating an audio file in the target language.

Architecture

"Architecture diagram"

Costs

This solution uses billable components of Google Cloud, including the following:

Use the pricing calculator to generate a cost estimate based on your projected usage.

Installation

  1. Prerequisites:

    • Python 3.7 or higher
    • Google Cloud SDK installed and configured
    • Google Cloud project with billing enabled

  2. Following Google Cloud services enabled:

    Use link Enable API or the command below:

    gcloud services enable \
speech.googleapis.com \
translate.googleapis.com \
texttospeech.googleapis.com

  3. Clone the repository and change to its directory

    cd projects/speech-to-speech_translation

    Note: it is recommended to use venv or pyenv virtualenv to create an isolated environment for this application and its imported modules. Please refer to the links in this step for more information on virtual environment

  4. Install dependencies:

    pip3 install -r requirements.txt --require-hashes

Usage example

To use the application, follow these steps:

  1. Create a configuration file:

    The application requires a configuration file config.ini in the application's main directory. You can use the below example as a starting point and replace the values with yours.

    [parameters]
project_id = your-project-ID
location = us-central1
gcs_path = gs://your-bucket/optional-path
stt_model = long
stt_timeout = 300
stt_alternative = 0
input_audio_file_name = input.wav
input_audio_file_path = your-server-local-path/
output_audio_file_name = output.wav
source_language_code = en-us
target_language = fi
target_language_code = fi-FI
target_voice = fi-FI-Wavenet-A
target_voice_gender = female
tts_timeout = 300
log = info
filename_prefix = timestamp

    Optionally, you can override the configuration file parameter values with command line switches of the same name. If both the configuration file and the command line arguments contain the same parameter, the command line argument's value will be used. You can execute the following command to see the command line arguments:

    python3 speech2speech.py --help

  2. Authenticate the shell:

    Execute the following command to authenticate your shell with the Google Cloud project in use:

    gcloud auth application-default login

  3. Run the application:

    After preparing the config.ini file, run the application with e.g:

    python3 speech2speech.py --output_interim_files

    The application will output its progress including the output files. Example:

    ...
2024-06-05 01:16:51,380 - root - INFO - Uploading input/en.wav to gs://bucket/path/en.wav
2024-06-05 01:16:53,690 - root - INFO - Waiting for STT operation to complete... timeout: 300s
2024-06-05 01:17:10,088 - root - INFO - Translating text to: fi
2024-06-05 01:17:11,051 - root - INFO - Waiting for TTS operation to complete... timeout: 300s
2024-06-05 01:17:28,616 - root - INFO - TTS output written to: gs://bucket/path/2024-06-05T01:16:49.593763_out.wav
2024-06-05 01:17:30,956 - root - INFO - Transcript written to: gs://bucket/path/2024-06-05T01:16:49.593763_transcript.txt
2024-06-05 01:17:32,982 - root - INFO - Translation written to: gs://bucket/path/2024-06-05T01:16:49.593763_translation.txt

Listing available Translation languages

To see the available languages supported by the Translation API, execute the following command:

python3 speech2speech.py --list_translate_languages

The command will output the supported languages as follows (snippet):

Name: Afrikaans
Language: af

Name: Albanian
Language: sq

Name: Amharic
Language: am
...

Listing available Text to Speech voices

To see the available voices and genders supported by the Text to Speech API, execute the following command:

python3 speech2speech.py --list_voices

The command will output the supported voice and gender options as follows (snippet):

Name: af-ZA-Standard-A
Supported language: af-ZA
SSML Voice Gender: FEMALE
Natural Sample Rate Hertz: 24000

Name: am-ET-Standard-A
Supported language: am-ET
SSML Voice Gender: FEMALE
Natural Sample Rate Hertz: 24000

Name: am-ET-Standard-B
Supported language: am-ET
SSML Voice Gender: MALE
Natural Sample Rate Hertz: 24000
...

Additional notes and limitations

  • The output quality of each interim stage is dependent on the AI API in use. E.g speech transcribing, language translation, and text to speech generation.
  • The data processing format (such as audio codec), duration, file size etc limitations are those of the respective API services used.
  • The stt_model parameter can be set to different Speech-to-Text models based on your needs. Please refer to the Speech-to-Text API documentation for details.
  • The target_voice parameter and target_voice_gender can be set to different Text-to-Speech voices based on the target language. Please refer to the Text-to-Speech API documentation for details.
  • The output_interim_files parameter allows you to save the transcribed text and translated text, in addition to the final audio output.

This user guide provides a comprehensive overview of the Speech-to-Speech Translation application. For more detailed information on Google Cloud services, refer to the official documentation.

License

This is not an Official Google Product.

Licensed under the Apache License, Version 2.0 (the "License")