Speech Settings

Configure NVDA’s speech output, synthesizers, and voice parameters

Overview

NVDA’s speech settings control how information is spoken to you. You can configure the synthesizer, voice characteristics, symbol pronunciation, and many other aspects of speech output.

Access Speech Settings: Press NVDA+control+s or navigate to Preferences > Settings > Speech in the NVDA menu.

Synthesizer Selection

The synthesizer is the text-to-speech engine that converts text into spoken output.

Changing the Synthesizer

1. Open Speech Settings

   Press NVDA+control+s to open the Speech category directly.

2. Click Change button

   In the Synthesizer section, click the Change… button.

3. Select a synthesizer

   Choose from available synthesizers in the list:

   • eSpeak NG (default, built-in, supports 80+ languages)
   • Windows OneCore voices (Windows 10/11 built-in voices)
   • Microsoft Speech API version 4 (SAPI 4)
   • Microsoft Speech API version 5 (SAPI 5)
   • Other synthesizers from add-ons
4. Click OK

   The new synthesizer will be activated immediately.

The Automatic setting lets NVDA choose the best available synthesizer. On Windows 10 and later, this is typically Windows OneCore voices.

Available Synthesizers

eSpeak NG (Default)

eSpeak NG is NVDA’s default synthesizer.

Advantages:

• Completely free and open source
• Supports over 80 languages
• Compact and fast
• Highly responsive
• No additional installation required
• Portable (works on any system)

Best for: General use, multiple languages, portable installations

Windows OneCore Voices

Windows OneCore voices are built into Windows 10 and later.

Advantages:

• Natural-sounding voices
• High-quality speech
• Integrated with Windows
• Supports many languages

Installation: Additional languages can be installed through Windows Settings > Time & Language > Language > Add a language.

Best for: Users who prefer more natural-sounding voices

Microsoft SAPI 5

SAPI 5 is a speech API that supports third-party voices.

Advantages:

• Compatible with commercial voices (e.g., Vocalizer, RealSpeak)
• Wide range of voice options
• Professional-quality voices available

Note: Requires separate installation of SAPI 5 voices.

Best for: Users with commercial voice licenses

Microsoft SAPI 4

SAPI 4 is an older speech API, still supported for compatibility.

Note: Requires installation of SAPI 4 voices (no longer included with Windows).

Best for: Legacy voice compatibility

Voice Settings

Once you’ve selected a synthesizer, you can configure voice parameters.

Voice Selection

Voice · dropdown

Select which voice to use from those available in the current synthesizer.

Different voices may have different:

• Genders (male, female, neutral)
• Ages (child, adult, elderly)
• Languages and accents
• Characteristics and tones

Voice Parameters

Rate · slider · default: "50"

Range: 0-100

Controls how fast the voice speaks. Higher values = faster speech.

• 0: Extremely slow
• 50: Normal speaking rate
• 100: Very fast

Keyboard shortcut to adjust:

• Increase: NVDA+control+upArrow
• Decrease: NVDA+control+downArrow

Pitch · slider · default: "50"

Range: 0-100

Controls the voice pitch (how high or low it sounds).

• 0: Very low pitch
• 50: Normal pitch
• 100: Very high pitch

Not all synthesizers support pitch adjustment.

Volume · slider · default: "100"

Range: 0-100

Controls speech volume.

• 0: Silent
• 100: Maximum volume

You can also adjust volume using NVDA+control+pageUp and NVDA+control+pageDown.

Inflection · slider · default: "50"

Range: 0-100

Controls voice inflection (variation in pitch during speech).

• 0: Monotone, flat speech
• 100: Highly expressive speech

Only available with some synthesizers (e.g., eSpeak NG).

Automatic Language Switching

Automatic language switching · boolean · default: "true"

When enabled, NVDA automatically changes the voice language when reading text marked with a different language.

Example: If you’re reading an English document with French quotes, NVDA will switch to a French voice for the quoted text.

Requires that the synthesizer supports the target language.

Automatic dialect switching · boolean · default: "false"

When enabled, NVDA attempts to switch between dialects (e.g., US English vs. UK English).

This is more specific than language switching and may not be supported by all synthesizers.

Trust voice's language · boolean · default: "true"

When enabled, NVDA trusts that the current voice can correctly handle the language of the text being read.

Symbol and Punctuation Settings

Symbol Level

Symbol level · dropdown · default: "Most"

Controls how much punctuation and symbols are announced.

Options:

• None: No punctuation is spoken
• Some: Only basic punctuation (period, comma, question mark, exclamation mark)
• Most: Most punctuation (adds semicolon, colon, brackets, etc.)
• All: All punctuation and symbols are spoken
• Character: Every character including letters is spoken with character descriptions

Keyboard shortcut to cycle: NVDA+p

For editing and proofreading, set symbol level to “All” or “Character”. For casual reading, “Most” or “Some” is usually sufficient.

Symbol Pronunciation

Customize how specific symbols are pronounced:

1. Click Symbol Pronunciation in the Speech Settings
2. Select a symbol from the list
3. Modify its:
   • Replacement: What is spoken for this symbol
   • Level: At which symbol level it should be announced
   • Send actual symbol: Whether to send the symbol itself to the synthesizer
4. Click OK to save

Symbol pronunciation is language-specific. Changes apply to the current interface language.

Capital Letter Announcements

Say 'cap' before capitals · boolean · default: "false"

When enabled, NVDA says “cap” before each capital letter.

Example: “Cap A” for the letter “A”

Beep for capitals · boolean · default: "false"

When enabled, NVDA plays a beep sound for capital letters.

Pitch change for capitals · slider · default: "30"

Range: -100 to +100

Changes voice pitch when speaking capital letters.

• Positive values: Raise pitch for capitals
• Negative values: Lower pitch for capitals
• 0: No pitch change

You can enable multiple capital announcement methods simultaneously.

Speech Modes

NVDA has multiple speech modes to control when and how speech output occurs:

Talk Mode (Default)

Normal speech mode.

NVDA speaks:

• Screen changes
• Notifications
• User actions
• Focus changes
• Everything as expected

On-Demand Mode

Speech only when requested.

NVDA only speaks when you:

• Use review commands
• Request information explicitly
• Press NVDA+tab, NVDA+t, etc.

Best for: Situations where you need occasional feedback but not continuous speech

Beeps Mode

Audio feedback instead of speech.

NVDA replaces speech with:

• Short beeps for events
• Different tones for different actions

Best for: Monitoring background activity, situations where speech is disruptive

Off Mode

No speech output.

NVDA is silent but still responds to commands.

Best for: Using only braille, giving presentations where speech would be distracting

Cycle through modes: Press NVDA+s

Excluded speech modes · multiselect

Select which speech modes should be skipped when cycling with NVDA+s.

Advanced Speech Settings

Character Descriptions

Delayed character descriptions · boolean · default: "false"

When enabled, NVDA waits briefly before providing character descriptions when typing, preventing them from interrupting your workflow.

Unicode Normalization

Unicode normalization · feature flag · default: "enabled"

Normalizes Unicode characters before sending to the synthesizer, which can improve speech quality for accented characters.

This is a feature flag that can be set to enabled, disabled, or default (which follows NVDA’s recommendation).

Report normalized for character navigation · boolean · default: "true"

When navigating by character, report the normalized form of characters.

Audio Processing

Trim leading silence · boolean · default: "true"

Removes silence at the beginning of speech output for faster response.

Only supported by some synthesizers.

Beep speech mode pitch · integer · default: "10000"

Range: 50-11025

Sets the pitch (in Hz) for beeps in Beeps speech mode.

Synth Settings Ring

The Synth Settings Ring allows you to quickly adjust common speech settings without opening the settings dialog:

Navigate Settings

• Next setting: NVDA+control+rightArrow
• Previous setting: NVDA+control+leftArrow

Adjust Values

• Increase: NVDA+control+upArrow
• Decrease: NVDA+control+downArrow
• Increase by larger steps: NVDA+control+pageUp
• Decrease by larger steps: NVDA+control+pageDown

Available settings in the ring:

• Voice
• Rate
• Pitch
• Volume
• Inflection (if supported)

The synth settings ring is extremely useful for making quick adjustments without interrupting your workflow.

Speech Dictionaries

Speech dictionaries allow you to customize pronunciation of specific words or patterns.

NVDA supports three dictionary levels:

Default Dictionary

Applies to: All voices and synthesizers

Contains pronunciations that should be used regardless of which voice or synthesizer is active.

Access: NVDA menu > Preferences > Speech dictionaries > Default dictionary

Voice Dictionary

Applies to: The current voice only

Contains pronunciations specific to the currently selected voice.

Access: NVDA menu > Preferences > Speech dictionaries > Voice dictionary

Temporary Dictionary

Applies to: Current NVDA session only

Contains pronunciations for the current session. Lost when NVDA restarts.

Access: NVDA menu > Preferences > Speech dictionaries > Temporary dictionary

Adding Dictionary Entries

1. Open a dictionary

   Select the appropriate dictionary from Speech dictionaries menu.

2. Click Add

   Opens the Add Dictionary Entry dialog.

3. Enter the pattern

   Type the word or pattern you want to affect in the Pattern field.

4. Enter the replacement

   Type how it should be pronounced in the Replacement field.

5. Set options

   • Case sensitive: Match only if case matches exactly
   • Regular expression: Use regex patterns for advanced matching
   • Comment: Optional note about this entry
6. Click OK

   Save the entry. It takes effect immediately.

Example: Make NVDA say “Doctor” instead of “Dr”:

• Pattern: Dr
• Replacement: Doctor
• Case sensitive: No

Troubleshooting Speech Issues

No speech output

Check:

1. Speech mode isn’t set to Off (press NVDA+s to cycle modes)
2. Volume isn’t at 0 (press NVDA+control+pageUp to increase)
3. Audio output device is working (test with other applications)
4. Synthesizer is properly configured (try switching to eSpeak NG)

Speech is too fast/slow

Adjust speech rate:

• Decrease rate: NVDA+control+downArrow
• Increase rate: NVDA+control+upArrow

Or open Speech Settings and adjust the Rate slider.

Voice sounds robotic or garbled

Try:

1. Switch to a different voice
2. Try a different synthesizer (Windows OneCore voices are more natural)
3. Adjust inflection settings (if available)
4. Check if Unicode normalization helps (in Speech Settings)

Wrong language being spoken

Check:

1. Correct voice language is selected
2. Automatic language switching is configured properly
3. Document language attributes are correct
4. Trust voice’s language setting is appropriate

Too much punctuation

Lower symbol level:

• Press NVDA+p to cycle to a lower level
• Or set it in Speech Settings

Recommended levels:

• Reading: Some or Most
• Editing: All

Related Topics

• NVDA Settings - General configuration
• Audio Settings - Audio output and ducking
• Voice Settings Panel - Detailed voice configuration
• Configuration Profiles - Different speech settings for different contexts