​

Speech Features

LibreChat supports both speech-to-text (STT) and text-to-speech (TTS) features, enabling voice-based interactions with AI models. You can speak your prompts and have responses read aloud.

​

Overview

Speech features in LibreChat include:

• Speech-to-Text (STT) - Convert spoken words to text input
• Text-to-Speech (TTS) - Convert AI responses to spoken audio
• Browser-based - Built-in browser APIs for basic functionality
• External engines - Connect to advanced speech services
• Conversation mode - Automatic hands-free back-and-forth interaction
• Voice selection - Choose from available voices
• Playback controls - Adjust speed and manage audio playback

​

Enabling Speech Features

​

Accessing Speech Settings

1. Open LibreChat settings (gear icon)
2. Navigate to the Speech tab
3. Configure STT and TTS options

​

Speech Settings Location

Settings > Speech contains all speech-related configuration options.

​

Speech-to-Text (STT)

Convert your spoken words into text for sending messages to the AI.

​

Enabling Speech-to-Text

1. Go to Settings > Speech > STT
2. Toggle Speech to Text switch to ON
3. The microphone button appears in the message input area

​

STT Engine Selection

Choose between available speech recognition engines:

​

Browser STT (Default)

Uses your browser’s built-in Web Speech API:

• Pros: No setup required, works offline (on some browsers), free
• Cons: Limited language support, accuracy varies by browser
• Best for: Quick testing, simple voice input, privacy-conscious users

​

External STT

Connects to external speech recognition services:

• Pros: Higher accuracy, more languages, better noise handling
• Cons: Requires configuration, may have costs, needs internet connection
• Best for: Production use, multilingual needs, professional applications

To select engine:

1. Go to Settings > Speech > STT
2. Find Engine dropdown
3. Select Browser or External
4. Save settings

​

Using Speech-to-Text

1. Ensure STT is enabled in settings
2. Click the microphone icon in the message input area
3. Allow microphone access if prompted by your browser
4. Start speaking
5. Your words appear as text in the input field in real-time
6. Click the microphone icon again to stop recording
7. Review and edit the transcribed text if needed
8. Send your message as usual

​

STT Visual Indicators

• Microphone icon - Default state, ready to record
• Red “Mic Off” icon - Currently recording
• Spinner - Processing speech (external STT)

​

Language Selection

For multilingual speech recognition:

1. Go to Settings > Speech > STT
2. Find Language dropdown
3. Select your preferred language
4. Available languages depend on your selected engine

Common languages:

• English (US, UK, Australian, etc.)
• Spanish
• French
• German
• Chinese
• Japanese
• And many more (varies by engine)

​

Advanced STT Settings

​

Auto-Transcribe Audio

Automatically transcribe when you finish speaking:

1. Go to Settings > Speech > STT
2. Toggle Auto-Transcribe Audio ON
3. Now when you stop speaking, transcription happens automatically

How it works:

• Detects when you stop speaking (based on silence)
• Automatically finalizes the transcription
• You don’t need to click the mic button to stop

​

Auto-Send Text

Automatically send transcribed messages:

1. Go to Settings > Speech > STT
2. Enable Auto-Send Text
3. Choose when to auto-send:
   • After transcription complete - Sends immediately when you stop speaking
   • On manual confirmation - Waits for you to confirm

Use cases:

• Hands-free conversation mode
• Quick voice queries
• Accessibility for users who can’t type

​

Decibel Threshold

Adjust microphone sensitivity:

1. Go to Settings > Speech > STT
2. Find Decibel Threshold slider
3. Lower values = more sensitive (picks up quieter sounds)
4. Higher values = less sensitive (requires louder speech)

Recommended settings:

• Quiet environment: -40 to -50 dB (mid-range)
• Noisy environment: -30 to -35 dB (less sensitive)
• Soft-spoken: -50 to -60 dB (more sensitive)

Test different levels to find what works best for your setup.

​

Text-to-Speech (TTS)

Have AI responses read aloud with natural-sounding voices.

​

Enabling Text-to-Speech

1. Go to Settings > Speech > TTS
2. Toggle Text to Speech switch to ON
3. Volume icons appear next to AI messages

​

TTS Engine Selection

​

Browser TTS (Default)

Uses your browser’s built-in speech synthesis:

• Pros: No setup, works offline, free, decent quality
• Cons: Limited voice selection, quality varies by browser/OS
• Best for: Basic TTS needs, testing, offline use

​

External TTS

Connects to external TTS services (e.g., OpenAI TTS, ElevenLabs, Google Cloud TTS):

• Pros: High-quality voices, natural intonation, emotion, more languages
• Cons: Requires API keys, may have costs, needs internet
• Best for: Professional use, high-quality voice output, specific voice needs

To select engine:

1. Go to Settings > Speech > TTS
2. Find Engine dropdown
3. Select Browser or External
4. If external, configure API credentials
5. Save settings

​

Using Text-to-Speech

1. Ensure TTS is enabled in settings
2. After the AI responds to your message
3. Click the volume icon next to the response
4. Audio playback begins
5. Click the icon again to stop playback

​

TTS Visual Indicators

• Volume icon - Ready to play, not currently playing
• Volume Mute icon - Currently playing (click to stop)
• Spinner - Loading/generating audio

​

Voice Selection

​

Browser Voices

1. Go to Settings > Speech > TTS
2. Find Voice dropdown
3. Select from available system voices
4. Voices vary by operating system:
   • Windows: Microsoft voices
   • macOS: Siri voices, others
   • Linux: eSpeak, Festival voices
   • Mobile: System-dependent

​

External Service Voices

1. Configure external TTS service
2. Go to Settings > Speech > TTS > Voice
3. Select from the service’s voice library
4. Options often include:
   • Multiple languages
   • Different genders
   • Various speaking styles (news, conversational, etc.)
   • Regional accents

​

Playback Controls

​

Playback Rate (Speed)

Adjust how fast the AI speaks:

1. Go to Settings > Speech > TTS
2. Find Playback Rate slider
3. Adjust speed:
   • 0.5x - Half speed (slower)
   • 1.0x - Normal speed
   • 2.0x - Double speed (faster)
   • Custom - Fine-tune to preference

Use cases:

• Slower (0.7x-0.9x): Learning new languages, complex topics
• Normal (1.0x): Standard listening
• Faster (1.2x-1.5x): Efficiency, familiar content

​

Global Audio Control

Manage playback across all messages:

• Pause all - Stop all currently playing audio
• Resume - Continue paused playback
• Skip - Move to next message in queue (if available)

​

Conversation Mode

Conversation mode enables continuous, hands-free voice interaction with the AI.

​

Enabling Conversation Mode

1. Go to Settings > Speech
2. Toggle Conversation Mode switch to ON
3. Both STT and TTS should be enabled

​

How Conversation Mode Works

1. You speak (STT transcribes)
2. Message auto-sends when you stop speaking
3. AI responds
4. Response is automatically read aloud (TTS)
5. After playback, mic activates for your next input
6. Cycle repeats for continuous conversation

​

Conversation Mode Requirements

• Speech-to-Text: Enabled
• Text-to-Speech: Enabled
• Auto-Transcribe: Recommended (auto-detects when you finish speaking)
• Auto-Send: Required (sends message automatically)

​

Starting a Conversation Mode Session

1. Enable conversation mode in settings
2. Start a new conversation or open existing one
3. Click the microphone icon
4. Begin speaking
5. The conversation flows automatically from there

​

Ending Conversation Mode

• Click the microphone icon to stop listening
• Turn off conversation mode in settings
• Close the conversation

​

Conversation Mode Best Practices

• Speak clearly - Better transcription accuracy
• Pause between thoughts - Helps auto-transcribe detect when you’re done
• Quiet environment - Reduces background noise interference
• Test settings first - Adjust decibel threshold and playback rate before long sessions

​

Tips and Best Practices

​

Speech-to-Text

• Test your microphone - Ensure it’s working properly before starting
• Check browser permissions - Allow microphone access when prompted
• Speak naturally - No need to speak robotically
• Review transcriptions - Edit any errors before sending
• Use punctuation commands - Some engines support saying “period”, “comma”, etc.
• Background noise - Minimize for better accuracy

​

Text-to-Speech

• Choose appropriate voice - Match language and preference
• Adjust playback speed - Find comfortable listening pace
• Use headphones - Better audio quality and privacy
• Long responses - Consider playback rate adjustment for lengthy content
• Pause and review - Stop playback to read complex code or data

​

Performance Optimization

• Browser vs External - External services often have better quality but require internet
• Disable when not needed - Turn off TTS/STT to save resources
• Cache voices - Some browsers cache voices for faster loading
• Test different engines - Compare quality and performance

​

Accessibility

• Vision impairment - TTS enables accessing AI responses without reading
• Motor impairment - Voice input as alternative to typing
• Dyslexia - Hearing responses can aid comprehension
• Multitasking - Listen to responses while doing other tasks

​

Troubleshooting

​

Speech-to-Text Issues

Microphone not working:

• Check browser permissions for microphone access
• Verify microphone is connected and not muted
• Try a different browser
• Check system audio settings

Poor transcription accuracy:

• Speak more clearly and slowly
• Reduce background noise
• Adjust decibel threshold
• Try external STT engine
• Select correct language in settings

No speech detected:

• Lower decibel threshold (increase sensitivity)
• Check microphone volume levels
• Ensure microphone isn’t muted or blocked
• Try speaking louder or closer to mic

Speech cuts off too early:

• Increase decibel threshold
• Disable auto-transcribe and stop manually
• Check browser STT settings

​

Text-to-Speech Issues

No audio playback:

• Check browser audio permissions
• Verify system volume isn’t muted
• Try a different voice or engine
• Check internet connection (for external TTS)
• Look for browser console errors

Voice sounds robotic:

• Try a different voice in settings
• Consider using external TTS service for higher quality
• Check if your browser has premium voices installed

Playback too fast/slow:

• Adjust playback rate in settings
• Reset to 1.0x if unsure
• Test incremental changes (0.1x adjustments)

Audio stuttering or lag:

• Check internet connection (for external TTS)
• Close other tabs using audio
• Try browser TTS instead of external
• Reduce browser resource usage

​

General Speech Issues

Settings not saving:

• Check browser local storage permissions
• Try refreshing the page
• Verify you’re logged in
• Clear browser cache if persistent

Feature not available:

• Speech features may be disabled by administrator
• Browser may not support Web Speech API
• External service may require configuration/API keys

​

Privacy and Security

​

Browser Speech APIs

• Processing happens on your device (generally)
• Some browsers may send audio to cloud services
• Check your browser’s privacy policy
• No API keys or external accounts needed

​

External Speech Services

• Audio sent to third-party service
• Subject to service provider’s privacy policy
• May be recorded or used for training
• Requires API credentials (stored securely)
• Consider sensitivity of content being spoken/read

​

Best Practices

• Review privacy policies of speech services you use
• Avoid speaking sensitive information if using cloud STT
• Use browser APIs for private/confidential conversations
• Disable speech features when not needed
• Check microphone permissions regularly

​

Keyboard Accessibility

Speech features are keyboard accessible:

• Tab - Navigate to microphone/speaker buttons
• Enter/Space - Activate speech recording or playback
• Escape - Stop recording or playback
• Navigate settings with keyboard

​

Advanced Configuration

​

Custom External Services

Administrators can configure custom STT/TTS endpoints:

• API endpoint URLs
• Authentication methods
• Request/response formats
• Voice/model selection

Check with your LibreChat administrator for available options.

​

Integration with Endpoints

Some LibreChat endpoints have native speech features:

• OpenAI’s Whisper (STT) and TTS
• Azure Speech Services
• Google Cloud Speech

These may offer tighter integration and better quality than generic external services.