Skip to main content

STT Settings

Configure Speech-to-Text (STT) to convert spoken audio into text for voice input. Choose between Chrome's built-in AI, OpenAI Whisper, or custom OpenAI-compatible endpoints.

Enable Speech-to-Text​

Toggle to enable or disable STT globally:

  • Enabled - Voice input available in chat
  • Disabled - Text-only input

Provider Selection​

VAssist supports three STT providers:

Chrome AI Multimodal​

On-device audio transcription using Chrome's AI

  • ✅ Free - No API costs
  • ✅ Private - All processing happens locally
  • ✅ Offline - Works without internet (after download)
  • ✅ No API Key - No account required
  • ⚠️ Requires Chrome 138+ with specific flags enabled
  • ⚠️ Limited Languages - English, Spanish, Japanese only

Best for: Privacy-conscious users, offline use, zero cost

OpenAI (Whisper)​

Cloud-based speech recognition using OpenAI's Whisper API

  • ✅ High Accuracy - Professional-grade transcription
  • ✅ Multi-Language - 50+ languages supported
  • ✅ Reliable - Production-ready service
  • ⚠️ Paid - Requires API key and usage billing

Best for: Multi-language support, production use, high accuracy needs

OpenAI-Compatible​

Custom STT endpoint compatible with OpenAI API

  • ✅ Flexible - Use any OpenAI-compatible STT service
  • ✅ Self-Hosted - Run your own Whisper server
  • ✅ Cost Control - No external API fees
  • ✅ Privacy - Data stays on your infrastructure

Best for: Custom deployments, self-hosted solutions, privacy

Chrome AI Multimodal Configuration​

Status & Availability​

Check if Chrome AI is available on your device:

  • Readily Available - Ready to use immediately
  • Downloadable - Click "Start Model Download" button
  • Downloading - Model downloading in progress
  • Unavailable - Not supported on this device

For real-time download progress, visit chrome://on-device-internals/

Click Refresh Status to recheck availability at any time.

Output Language​

Options: English (en), Spanish (es), Japanese (ja)
Default: English (en)

Language for transcription output. Chrome AI currently supports only these three languages.

info

Chrome AI uses the same model as the LLM provider. If you already have it downloaded for LLM, you don't need to download again.

Required Chrome Flags​

Chrome AI STT requires these flags to be enabled:

  1. optimization-guide-on-device-model

    • Set to: Enabled BypassPerfRequirement

  2. prompt-api-for-gemini-nano-multimodal-input

    • Set to: Enabled

After enabling flags:

  1. Restart Chrome
  2. Visit chrome://components
  3. Find "Optimization Guide On Device Model"
  4. Click "Check for update" to download the AI model
tip

These are the same flags required for Chrome AI LLM provider. If you've already enabled them, you're good to go!

OpenAI (Whisper) Configuration​

API Key​

Required

Your OpenAI API key from platform.openai.com

Format: sk-...

tip

Keep your API key secure. VAssist stores it locally in browser storage.

Model​

Default: whisper-1

The OpenAI Whisper model to use for transcription.

Currently, OpenAI only offers whisper-1 model. Leave as default unless OpenAI releases new models.

Features​

OpenAI Whisper provides:

  • Multi-Language Support - 50+ languages automatically detected
  • High Accuracy - State-of-the-art speech recognition
  • Punctuation - Automatic punctuation in transcripts
  • Timestamps - Optional word-level timestamps

See OpenAI Whisper API for details.

OpenAI-Compatible Configuration​

For self-hosted or custom STT services that implement the OpenAI API format.

Endpoint URL​

Required

Base URL of your STT service. VAssist will append /v1/audio/transcriptions automatically.

API Key (Optional)​

Leave empty if your endpoint doesn't require authentication.

Enter the API key if your service requires it for access control.

Model​

Default: whisper

Model identifier your service expects. Depends on your STT implementation.

Consult your STT provider's documentation for supported models.

Testing STT​

Click Test Recording (3s) to verify your configuration:

  1. Grant Microphone Permission - Browser will request access
  2. Speak for 3 seconds - Recording starts automatically
  3. Wait for Transcription - Processing may take a few seconds
  4. Check Result - Transcribed text appears in alert/console

Test requirements:

  • STT must be enabled
  • Provider must be configured correctly
  • Microphone access granted
  • For Chrome AI: Model must be downloaded
  • For OpenAI: Valid API key and credits
  • For OpenAI-Compatible: Endpoint must be reachable
info

The test records for exactly 3 seconds to quickly verify the setup. Real voice input in chat can record for up to 60 seconds.

Tips​

For Privacy​

  • Chrome AI: 100% on-device, no data sent anywhere
  • OpenAI-Compatible: Self-hosted option keeps data local

Troubleshooting​

Chrome AI Not Available​

  • Update Chrome to version 138 or later
  • Enable required flags (see above)
  • Check device compatibility at chrome://on-device-internals/
  • Restart Chrome after enabling flags

Microphone Not Working​

  • Check browser permissions (allow microphone access)
  • Verify microphone is working in system settings
  • Try different browser/profile
  • Check if another app is using the microphone

No Transcription / Empty Result​

  • Chrome AI:
    • Verify model is downloaded (check status)
    • Try switching to English language
    • Speak louder/clearer
  • OpenAI:
    • Verify API key is correct
    • Check you have credits in OpenAI account
    • Ensure audio is audible
  • OpenAI-Compatible:
    • Verify endpoint is reachable
    • Check server logs for errors
    • Ensure model name is correct

Poor Transcription Quality​

  • Use better microphone (not built-in laptop mic)
  • Reduce background noise
  • Speak clearly and at normal pace
  • Check microphone input level in system settings

Next Steps​