Voice Dictation with Sarvam AI
Guide for setting up and using Sarvam AI for voice dictation and translation.
Voice Dictation with Sarvam AI
Overview
Voice dictation enables you to communicate with Sypha through speech rather than typing. This capability now includes Sarvam AI, a transcription service specialized in Indian languages, in addition to the existing Sypha provider.
Features
- Speech-to-Text Transcription: Transform your spoken words into written text
- Multiple Languages: Compatibility with English and 10+ Indian languages
- Translation: Automatic translation of transcribed content to a different language
- Provider Choice: Option to select between Sypha (deprecated) and Sarvam AI
- Secure API Key Storage: Your credentials are securely maintained in VSCode
Supported Languages (Sarvam AI)
Sarvam AI focuses on Indian languages and provides support for:
- English (India)
- Hindi (हिन्दी)
- Bengali (বাংলা)
- Gujarati (ગુજરાતી)
- Kannada (ಕನ್ನಡ)
- Malayalam (മലയാളം)
- Marathi (मराठी)
- Odia (ଓଡ଼ିଆ)
- Punjabi (ਪੰਜਾਬੀ)
- Tamil (தமிழ்)
- Telugu (తెలుగు)
Getting Started
1. Get a Sarvam AI API Key
- Navigate to Sarvam AI website
- Register for an account
- Access your dashboard
- Create an API key for speech-to-text services
- Copy your API key (required for the following step)
2. Configure Sypha for Sarvam AI
- Access Sypha settings (gear icon in the chat interface)
- Go to General Settings > Features
- Activate Dictation checkbox
- From the provider dropdown, choose "Sarvam AI (Indian Languages)"
- Input your Sarvam AI API key in the secure input field
- Choose your Transcription Language (the language you will speak)
- (Optional) Activate Translation if you wish to translate your speech to a different language
- Save settings
3. Start Using Voice Dictation
- Within the chat input, select the microphone icon 🎤
- Articulate your message clearly
- Select the stop button upon completion
- The transcribed (and optionally translated) text will populate the input box
- Review and modify if necessary
- Submit your message
Translation Feature
The translation capability enables you to speak in one language and have it automatically converted to another prior to submission to the AI.
How to Use Translation
- Activate dictation and configure Sarvam AI as outlined above
- Activate the "Enable Translation" checkbox
- Choose your target language from the "Translate To" dropdown
- Begin dictating - your speech will undergo:
- Initial transcription in your spoken language
- Subsequent translation to your target language
- Finally, the translated text displays in the input box
Example Use Cases
- Speak Hindi, Send English: Articulate in Hindi, translate to English for the AI
- Speak Tamil, Send Hindi: Articulate in Tamil, translate to Hindi
- Speak Kannada, Send English: Articulate in Kannada, translate to English
Settings Reference
Transcription Provider
- Sypha (Deprecated): Legacy provider, requires Sypha account. Will be discontinued in future versions.
- Sarvam AI: Contemporary provider with exceptional Indian language support. Recommended.
Transcription Language
The language you will articulate when utilizing voice dictation. Sarvam AI will transcribe your speech in this language.
Enable Translation
Toggle to activate automatic translation of transcribed content.
Translation Target Language
The language to convert your speech into prior to submission to the AI assistant.
Troubleshooting
"Invalid Sarvam AI API key"
Solution:
- Confirm your API key is accurate
- Verify you copied the complete key without spaces
- Confirm that your API key possesses permissions for speech-to-text
- Create a new API key if necessary
"Cannot connect to Sarvam AI"
Solution:
- Verify your internet connection
- Confirm Sarvam AI service is operational
- Attempt again momentarily
"Unsupported language"
Solution:
- Confirm you've chosen a language from the supported list
- Sarvam AI presently supports exclusively Indian languages and English (India)
- If you require other languages, utilize the Sypha provider
"Rate limit exceeded"
Solution:
- Wait briefly before attempting again
- Verify your Sarvam AI account quota
- Upgrade your Sarvam AI plan if necessary
"Translation failed"
Solution:
- Original transcription is utilized when translation fails
- Verify your API key possesses translation permissions
- Confirm both source and target languages are supported
- Attempt recording again
Transcription is inaccurate
Tips for better transcription:
- Articulate clearly and at a moderate pace
- Utilize a quality microphone
- Minimize background noise
- Articulate in complete sentences
- If transcription is consistently inadequate, attempt a different provider
Privacy and Security
- API Key Storage: Your Sarvam AI API key is maintained securely in VSCode's encrypted secrets storage
- Audio Processing: Audio is transmitted directly to Sarvam AI servers for processing
- No Recording Storage: Sypha does not retain your voice recordings
- Temporary Files: Audio files are automatically removed following transcription
Pricing
Sarvam AI pricing differs by usage. Please review Sarvam AI's pricing page for current rates.
Note: Sypha does not impose charges for utilizing Sarvam AI - you compensate Sarvam AI directly according to your usage.
Comparison: Sypha vs Sarvam AI
| Feature | Sypha (Deprecated) | Sarvam AI |
|---|---|---|
| Requires Account | ✅ Yes (Sypha) | ❌ No (just API key) |
| Indian Languages | ❌ Limited | ✅ Excellent (10+ languages) |
| Translation | ❌ No | ✅ Yes |
| Future Support | ❌ Being deprecated | ✅ Actively supported |
| Pricing | Sypha credits | Pay-as-you-go to Sarvam |
Migration from Sypha to Sarvam AI
If you're presently utilizing Sypha for dictation:
- Obtain a Sarvam AI API key (see "Getting Started")
- Access Sypha settings
- Modify provider from "Sypha" to "Sarvam AI"
- Input your Sarvam AI API key
- Test with a brief recording
- You're prepared!
Note: Your existing Sypha transcriptions and settings will remain unaffected.
Best Practices
- Test First: Record a brief test message to confirm setup
- Review Transcriptions: Always examine transcribed text prior to submission
- Clear Speech: Articulate clearly for optimal results
- Appropriate Length: Maintain recordings under 2-3 minutes for optimal performance
- Stable Internet: Guarantee stable internet connection throughout transcription
- API Key Security: Never disclose your API key publicly
FAQ
Q: Can I use both Sypha and Sarvam AI?
A: You can alternate between providers in settings, but exclusively one can be active simultaneously.
Q: Does translation work with Sypha provider?
A: No, translation is presently available exclusively with Sarvam AI.
Q: How long can my recordings be?
A: Maximum 5 minutes per recording, but shorter recordings (1-2 minutes) typically perform optimally.
Q: Can I use Sarvam AI for languages other than Indian languages?
A: Sarvam AI presently specializes in Indian languages and English (India). For other languages, you may need to utilize a different provider.
Q: Is my voice data stored anywhere?
A: Sypha does not retain your voice recordings. They are transmitted to Sarvam AI for processing and subsequently immediately removed. Review Sarvam AI's privacy policy for their data handling practices.
Q: Can I cancel a recording in progress?
A: Yes, select the X button that displays during recording to cancel.
Support
If you experience issues:
- Review this documentation for solutions
- Confirm your API key and internet connection
- Attempt the Sypha provider to isolate the issue
- Review Sarvam AI status page
- Contact Sarvam AI support for API-related issues
- Report bugs to Sypha development team
Feedback
We're continuously enhancing voice dictation. Share your feedback:
- Report issues on GitHub
- Propose new features
- Share your use cases
- Assist us in improving Indian language support
Last Updated: [Current Date] Version: 1.0.0 Provider: Sarvam AI