WorldVoice User Guide

WorldVoice is an NVDA add-on that supports automatic switching between speech roles from different speech engines, including Espeak, OneCore and SAPI5, and provides a wide range of customization settings. Users can easily adjust speech settings for different situations and personal preferences.

As globalization becomes more widespread, people are exposed to multilingual written content more often. Whether in educational material or leisure reading, it is common to see different languages mixed together, and sometimes Chinese, English, and Japanese text can even appear in the same sentence.

Different types of digital content, such as language-learning material, mathematical information, or literary works, vary in text structure, context, and frequency of language mixing. When using a screen reader, being able to adjust reading behavior flexibly according to the content and personal preference can improve reading quality.

Features

• Assign dedicated speech roles to different languages, so multilingual content can automatically switch to a more suitable voice while being read.
• Configure rate, pitch, and volume independently for each speech role, so different languages can keep their own reading settings.
• Switch number reading modes as needed, choosing whether numbers are read as values or digit by digit.
• Adjust pause lengths between numbers, items, Chinese text, and say-all reading, so the reading rhythm better matches user preference.
• Detect language automatically from Unicode characters, reducing the need to switch voices manually when reading multilingual content.

Speech Roles

In the WorldVoice speech role settings panel, you can assign dedicated speech roles to different languages and adjust parameters such as rate, pitch, and volume for each role individually.

WorldVoice stores settings for each speech role separately. When a role is switched, the corresponding settings are applied, allowing each speech role to keep its own rate, pitch, and volume. This is especially helpful when reading content in a non-native language.

• After selecting a language or region, the voice list shows the speech roles available for that language or region. Selecting a speech role completes the mapping between that language or region and the speech role.
• After selecting a speech role, if that role supports voice variants, the available variant options are shown for selection.
• After a speech role is selected, the rate, pitch, and volume sliders, as well as the rate boost checkbox, automatically update to that role's settings. Changing these values affects only the currently selected speech role, because each speech role has its own independent settings.

Consistency Settings

• Keep the main speech engine and regional speech engines consistent: When enabled, the main speech role and regional speech roles can only use speech roles from the same speech engine. If the main speech role is changed and a regional speech role belongs to a different speech engine, that regional speech role is reset to an unselected state and can only be selected again from the same engine as the main speech role.
• Keep the main speech role and regional speech roles consistent: When enabled, the main speech role and regional speech roles remain the same. When you change the speech role setting on either side, the other side is automatically synchronized to the same speech role.
• Keep parameters consistent between different speech roles: When enabled, rate, pitch, volume, and rate boost parameters remain consistent across all speech roles. Changing the settings for any speech role updates the settings for all other speech roles as well.

The main speech role is configured through the NVDA voice settings panel, while regional speech roles are configured through the WorldVoice speech role settings panel.

Speech Pipeline

Speech pipeline settings control reading behavior related to numbers, pauses, and language processing. You can choose whether these settings apply to all supported synthesizers or only within WorldVoice.

• Globally supported speech pipeline settings: ignore commas between numbers, number mode, item pause, number pause, Chinese pause, say-all pause

• WorldVoice-only speech pipeline settings: detect language based on Unicode characters, number language

• Detect language based on Unicode characters: When enabled, WorldVoice automatically determines the language or region from the text's Unicode characters and switches speech roles accordingly. Note: This feature may conflict with NVDA's automatic language switching, so it is recommended not to enable both at the same time.

• Number language: When reading numbers, use the regional role configured by this option.
• Number mode:
• Numeric mode: Reads numbers as values. For example, "12345" is read as "twelve thousand three hundred forty-five."
• Digit mode: Reads numbers digit by digit. For example, "12345" is read as "one two three four five."
• Speech pause adjustment: Sets the pause length between numbers, items, Chinese text, and say-all reading. Smaller values mean shorter pauses, and 0 means no pause.
• Ignore commas between numbers: Ignores commas between digits when reading numbers to improve number-reading accuracy for specific speech roles.

Globally supported speech pipeline settings can be adjusted through the NVDA speech settings panel or the WorldVoice speech pipeline panel. Speech pipeline settings that are not globally supported can only be adjusted through the NVDA speech settings panel.

Unicode Detection

These options adjust how WorldVoice handles numbers, punctuation, and character sets shared by multiple languages when detecting text language.

• Ignore numbers during language detection: When checked, numbers are ignored during detection and are read using the current speech role.
• Ignore common punctuation during language detection: When checked, punctuation is ignored during detection and is read using the current speech role.
• Language detection timing: Determines whether Unicode-based language detection and language-switching commands are handled before or after NVDA processes speech commands.
• Assumed language for character sets: Sets the default language for Unicode character sets shared by multiple languages, including Latin characters, CJK characters, and Arabic characters.

Notes

• If you enable both NVDA automatic language switching and WorldVoice Unicode detection, the two features may affect each other. Choose one according to your needs.
• If you want to disable a specific speech engine, clear the corresponding checkbox in the WorldVoice speech engine panel.
• If you need to use a speech engine other than Espeak, OneCore, or SAPI5, download the corresponding driver or voice package separately.

WorldVoice Changelog

WorldVoice 6 Update

• Fixed responsiveness during delayed descriptions: Break tasks are now cancellable, reducing lag when delayed character descriptions are enabled.
• Improved speech dictionary compatibility: Speech dictionaries are now applied before the WorldVoice pipeline.
• Refined pipeline settings internals: Pipeline settings handling has been refactored for clearer behavior and easier maintenance.
• Improved voice metadata handling: Duplicate voice names across engines are now disambiguated, and fallback language descriptions are normalized.
• Refactored driver architecture: Engine driver architecture and loading have been redesigned, replacing hard-coded loading with manifest-based discovery.
• Refined core packaging: Core driver packaging has been reorganized.
• Improved Sonic initialization: Sonic is now initialized lazily on first voice callback.
• Refined task execution internals: TaskManager now uses a Future-based design.
• Improved built-in engine integration: SAPI5, OneCore, and eSpeak now inherit from the built-in base class.
• Improved number pronunciation logic: Number pronunciation and translation logic have been refined.

WorldVoice 5 Update

• Added features and settings panels: Speech Pipeline and Log Record
• Removed deprecated features and settings panels: Unicode normalization, number-dot replacement, and other outdated settings
• Added speech pipeline feature: Enable WorldVoice's speech pipeline for all synthesizers
• Added log recording feature: Capture speech sequences before and after each pipeline step for debugging
• Added rate boost feature: Allow rate boost for specific speech engines
• Improved user experience: Improved interactions in the Speech Engine and Log Record dialogs
• Revised interface text: Updated interface text for a consistent style

WorldVoice 4 Update

• Integrated the Espeak engine into the engines supported by WorldVoice.
• Added a rate boost setting to the NVDA voice settings panel.
• Added a voice variant setting to the NVDA voice settings panel.
• If the current engine or voice supports rate boost, the NVDA voice settings panel dynamically displays the rate boost setting UI.
• Users can set rate boost for individual voices in the WorldVoice voice settings panel.