Skip to main content

Google Speech

The Google Speech package provides powerful cloud-based speech recognition and synthesis capabilities, enabling you to convert speech to text and text to speech in your RPA workflows using Google Cloud's advanced AI models.

Overview​

Google Cloud Speech services offer state-of-the-art machine learning models for audio processing. This package integrates both Speech-to-Text (STT) and Text-to-Speech (TTS) capabilities, allowing you to build sophisticated voice-enabled automation workflows.

Key Features​

Speech Recognition​

  • Convert audio files to text with high accuracy
  • Support for multiple languages and dialects
  • Automatic punctuation and formatting
  • Word-level timestamps for precise audio alignment
  • Multiple recognition models optimized for different audio sources
  • Long-running recognition for audio files of any length

Speech Synthesis​

  • Generate natural-sounding speech from text
  • Wide selection of voices across multiple languages
  • Multiple audio formats (WAV, MP3, OGG Opus)
  • SSML support for advanced speech control
  • Customizable voice parameters (pitch, speed, volume)
  • Studio-quality voices with neural network technology

Authentication​

Both nodes require Google Cloud service account credentials with the appropriate API permissions:

  1. Speech-to-Text API: For audio transcription
  2. Text-to-Speech API: For speech synthesis

Your service account must have the necessary roles (e.g., "Cloud Speech Client") and the APIs must be enabled in your Google Cloud project.

Common Use Cases​

Customer Service Automation​

  • Transcribe customer call recordings for analysis
  • Generate automated voice responses and notifications
  • Create voice-based IVR systems
  • Process voicemail messages automatically

Content Creation​

  • Generate voiceovers for videos and presentations
  • Create audio versions of written content
  • Produce multilingual audio content
  • Build accessible applications with voice output

Data Processing​

  • Extract data from voice recordings and interviews
  • Transcribe meeting recordings for documentation
  • Process audio surveys and feedback
  • Create searchable transcripts from audio archives

Accessibility​

  • Convert written content to speech for visually impaired users
  • Transcribe audio content for hearing impaired users
  • Build voice-controlled automation workflows
  • Create multilingual voice interfaces

Available Nodes​

Speech Recognition​

  • Speech to Text: Convert audio files from Google Cloud Storage to text with advanced recognition features

Speech Synthesis​

  • Text to Speech: Convert text or SSML to natural-sounding speech audio files

Getting Started​

  1. Set up Google Cloud:

    • Create a Google Cloud project
    • Enable Speech-to-Text and/or Text-to-Speech APIs
    • Create a service account with appropriate permissions
    • Download the service account JSON key file

  2. Configure Credentials:

    • Add your service account credentials to Robomotion
    • Use the credentials in the Speech nodes

  3. For Speech-to-Text:

    • Upload your audio file to Google Cloud Storage
    • Use the Speech to Text node with the GCS URI
    • Receive transcription results with confidence scores

  4. For Text-to-Speech:

    • Prepare your text content (plain or SSML)
    • Select voice and audio format
    • Generate speech and save to file

Supported Languages​

Google Speech supports 100+ languages and variants including:

  • English (US, UK, AU, IN, etc.)
  • Spanish (ES, MX, US, etc.)
  • French (FR, CA)
  • German (DE)
  • Chinese (Mandarin, Cantonese)
  • Japanese
  • Korean
  • Arabic
  • Hindi
  • Portuguese (BR, PT)
  • And many more...

See the Google Cloud Speech documentation for the complete list of supported languages.

Audio Format Requirements​

Speech-to-Text​

  • Audio must be stored in Google Cloud Storage
  • Supported formats: WAV (LINEAR16), FLAC, MP3, OGG Opus, AMR, SPEEX
  • Recommended: WAV with 16kHz sample rate for best results
  • Maximum audio length: 480 minutes

Text-to-Speech​

  • Output formats: WAV (LINEAR16), MP3, OGG Opus
  • Configurable sample rates (8kHz to 48kHz)
  • Default: 16kHz for optimal quality and file size

Pricing Considerations​

Google Cloud Speech services are billed based on usage:

  • Speech-to-Text: Charged per 15 seconds of audio processed
  • Text-to-Speech: Charged per 1 million characters processed

Different pricing tiers apply for:

  • Standard vs. premium voices (TTS)
  • Different recognition models (STT)
  • Enhanced models and features

See Google Cloud Pricing for detailed information.

Best Practices​

For Speech Recognition​

  • Use appropriate audio sample rates (16kHz recommended)
  • Store audio files in Google Cloud Storage for best performance
  • Select the correct recognition model for your audio source
  • Enable automatic punctuation for better readability
  • Use language-specific models for improved accuracy
  • Consider using enhanced models for noisy audio

For Speech Synthesis​

  • Use SSML for fine-grained control over pronunciation and timing
  • Select voices that match your target audience and use case
  • Choose appropriate audio formats based on quality vs. size needs
  • Test different voices to find the most natural-sounding option
  • Use Studio voices for highest quality output
  • Cache generated audio files to reduce API calls

General​

  • Implement proper error handling for network issues
  • Monitor API usage and costs
  • Use regional endpoints when possible for lower latency
  • Validate credentials before processing large batches
  • Consider rate limits and quotas in production workflows

Tips for Effective Use​

  1. Optimize Audio Quality: Higher quality input audio produces better transcription results
  2. Use Appropriate Models: Select models based on your audio source (video, phone, default)
  3. Language Selection: Always specify the correct language code for best accuracy
  4. Batch Processing: Process multiple files efficiently by reusing credentials
  5. Error Recovery: Implement retry logic for transient network errors
  6. Voice Selection: Test multiple voices to find the best fit for your content
  7. SSML Formatting: Use SSML tags to control emphasis, pauses, and pronunciation
  8. File Management: Clean up temporary audio files after processing

Limitations and Quotas​

Speech-to-Text​

  • Maximum audio length: 480 minutes (8 hours)
  • Long-running recognition for files over 1 minute
  • Synchronous recognition limited to 1 minute audio
  • Rate limits apply per project

Text-to-Speech​

  • Maximum input text: 5,000 characters per request
  • Maximum SSML length: 5,000 characters
  • Rate limits apply per project
  • Some voices have usage restrictions

Common Errors and Solutions​

Authentication Errors​

  • Invalid credentials: Verify JSON key file is complete and valid
  • Permission denied: Ensure service account has required API roles
  • API not enabled: Enable Speech-to-Text/Text-to-Speech API in Cloud Console

Speech-to-Text Errors​

  • Invalid GCS URI: Verify the gs:// path is correct and accessible
  • Unsupported audio format: Convert audio to supported format
  • Sample rate mismatch: Ensure configured rate matches actual audio
  • File not found: Verify service account has Storage Object Viewer role

Text-to-Speech Errors​

  • Invalid voice name: Check voice is available for selected language
  • Text too long: Split large texts into smaller chunks
  • Invalid language code: Use correct BCP-47 language codes
  • File write error: Verify output path is writable

Resources​