Rehnuma Awaz Updates

The Rehnuma Awaz NVDA add-on has been rebuilt from the ground up for faster speech, clearer Urdu pronunciation, and smooth switching between Urdu and English.

• Clearer Urdu reading: an expanded 18,000+ word dictionary, natural phrasing for long sentences, and improved handling of numbers, dates, punctuation, and Quranic symbols. Short English phrases can now be read directly by the Urdu voice.
• Character and symbol processing: support for all Urdu characters while typing or moving character by character, and the Rehnuma Symbols dictionary to help you control symbol announcements via NVDA’s symbol level.
• Faster performance: speech starts up to 40% sooner, uses half the CPU power, and eliminates audio crackling and cut-offs at high speeds.
• Seamless audio switching: Urdu and English share a single audio stream to eliminate gaps and overlapping. Fully compatible with third-party auto-switchers like Multilang.
• Flexible voice options: choose from multiple Urdu and English voices, or pair Rehnuma Awaz with other NVDA synthesizers (e.g. Windows OneCore, IBMTTS, eSpeak NG) without third-party switchers.
• New Voice Manager: download, preview, update, and manage voices from NVDA’s Tools menu (NVDA+Shift+Alt+R).
• Rehnuma settings panel: adjust sentence and comma pause durations, voice rhythm, and choose whether numbers are read in Urdu, English, or the surrounding language.
• Urdu sentence navigation: extended NVDA’s sentence navigation for Microsoft Word to support Urdu sentences as well.
• Auto-updates: built-in checks keep the add-on up to date without deleting downloaded voices.
• Comprehensive documentation: detailed user guides in both Urdu and English, available from NVDA’s Help menu.

Enhancements: adds full support for NVDA 2026.1, restores Rehnuma speech and the Voice Manager after NVDA updates, and remains compatible with NVDA 2024–2025.

Restart NVDA after installing.

Improved Urdu character recognition: special characters and diacritics such as superscript alef (ٰ), do-chashmi heh (ھ), and other Urdu-specific marks are now correctly recognized as Urdu text instead of fallback “unknown” or “Arabic letter” tokens.

Numeric handling update: Arabic numerals (0–9) are now automatically treated as part of Urdu text, ensuring consistent pronunciation and TTS handling of numbers within Urdu sentences.

Impact: TTS output for Urdu text with diacritics and embedded numbers is now more accurate and natural.

• Enhanced restart flow after new voice installation: fixed issues causing delayed or incorrect restarts when adding new voices. The system now restarts cleanly and reliably after each addition.
• Smarter auto-language switching: improved detection and switching logic for multilingual content, resulting in smoother transitions and fewer misclassifications.

Rehnuma Manager UI upgrade

• Polished interface for better usability and workflow.
• Added support for registering other Piper-based English voices and models.
• Introduced quick voice selection directly inside the Rehnuma Voice Manager panel.

Voice improvements

• Rehnuma Akbar: significantly improved handling of single words and individual letters, making it more reliable for TTS training, dictionary work, and educational use.
• New voice, Rehnuma Arfa (female): a natural-sounding female voice named Arfa to broaden the selection for users.

Miscellaneous

• Various stability fixes, performance tuning, refactoring, and internal improvements.

• Enabled ONNX Runtime graph optimizations
• Set GraphOptimizationLevel to Level3 and enabled memory pattern reuse for inference sessions
• Impact: 15–45% faster model inference; fewer allocations during repeated calls
• Added phonemization result caching
• Thread-safe cache keyed by voice and processed text (includes Arabic diacritized text)
• Impact: 50–90% lower latency for repeated or similar utterances; lower CPU usage
• Fixed exponential growth in real-time streaming chunk size
• Replaced multiplicative chunk size update with constant base chunk size per stream
• Impact: predictable latency, reduced memory usage, prevents performance degradation over time
• Enabled GPU acceleration on Windows via DirectML
• ONNX Runtime provider ordering: DirectML (preferred) → CPU fallback
• Impact: large speedups on compatible GPUs without configuration changes
• Tuned real-time streaming parameters
• Updated defaults: chunk_size=64, chunk_padding=2
• Impact: lower initial latency while maintaining stream smoothness
• Increased gRPC channel buffer sizes
• mpsc channel capacity raised from 512 to 1024 for both synthesis endpoints
• Impact: reduced backpressure and blocking under bursty load; smoother streaming
• Configured gRPC server for high concurrency
• Set concurrency_limit_per_connection=1024, max_concurrent_streams=1000, tcp_keepalive=60s, tcp_nodelay=true
• Impact: improved throughput and responsiveness under concurrent client load
• Right-sized synthesis thread pool
• Reduced from num_cpus×4 to min(num_cpus×2, 16), with a floor of 2
• Impact: less context switching, more stable latency, better CPU utilization

• Improved number reading for more accurate and natural pronunciation.
• Enhanced the voice download page with better accessibility and easier navigation.

Release notes

• Read punctuation is now a selection screen with four levels: None, Some, Most, and All (replacing the on/off switch).
• Number processing is now a selection screen with: Full for all digits, Full for pairs only (below 99), Full for triplets only (below 999), and Individual for all digits.
• Time announcement is now a selection screen with: English, Urdu full (natural phrasing with period of day), and Urdu short (e.g. 2:30 → “دو تیس”).
• Other settings: a new screen for audio ducking, enable logging, and share crash report (moved out of the main screen and Technical Details).
• Speech settings (Urdu and English): a “Jump by 5” toggle so sliders move in steps of 5 for easier navigation with TalkBack; off keeps 1-step movement. Removed the number above each slider (TalkBack already announces the value when the slider is focused).
• Accessibility and reliability: the Time Announcement, Read Punctuation, and Number Processing screens now announce the correct screen title in TalkBack (not “Voice Selection”).
• Voice profile selection no longer shows a false “Failed” message when the voice was saved successfully.
• Branding: replaced the Tech Destination and old MoITT partner logos with the Ministry of IT & Telecom logo on the main and About screens.
• Default voice improved.

• Screen reader stability: optimized engine thread synchronization to prevent speech pauses and interruptions under rapid back-to-back requests (especially when using active screen readers like Jieshuo or TalkBack).
• Fixed speech speed issues: resolved a race condition that could cause speech speeds to apply incorrectly during rapid English–Urdu language switching.
• Accurate debug profiler: telemetry reports now accurately record the active voice model per segment and show package names when external engines are active.
• General performance optimizations and bug fixes.

Release notes

• Removed unnecessary text from the Home screen for a cleaner experience.
• Improved speech generation for very long text passages, making playback more reliable.
• Fixed the audio ducking issue.
• Added an option to announce numbers in Urdu.

This update brings more customization, performance improvements, and audio tuning controls:

• Advanced mode improvements: set preferred speech engines separately for Urdu and English, for finer control over voice output.
• Time announcement fixes: improved accuracy and reliability of time announcements.
• New pitch control slider: adjust the voice pitch to match your listening preference.
• Dynamic core allocation: smarter CPU core usage based on the device, for smoother and more efficient playback.
• General stability improvements and bug fixes.

Release notes

• Increased maximum voice speed up to 6× for faster listening.
• Added separate voice speed controls for Urdu and English voices.
• Added separate volume controls for Urdu and English voices.
• Choose whether time announcements are spoken in Urdu or English.
• Improved Urdu time announcements for greater accuracy and natural pronunciation.
• Added an audio ducking toggle to control whether other media volumes are lowered during announcements.
• New option to announce numbers individually or as complete numbers.
• Various stability improvements, bug fixes, and performance enhancements.

• Works with AutoTTS, which switches voices by language for you.

• Added an in-app user guide for better accessibility.
• Removed outdated voices.
• Improved overall UI/UX for smoother navigation.

• Performance optimizations for faster and more stable operation.

• Added 3 new English voices and 1 new Urdu voice.

• Enabled compatibility with Voiceify integration.

• Increased speech speed rate for improved reading efficiency.

• Improved word-by-word and character-by-character reading accuracy.

• iOS 27 compatibility: improved compatibility with iOS 27.
• Urdu auto-switching: fixed an issue affecting automatic language switching when reading Urdu text.

• iOS 27 compatibility: improved compatibility with iOS 27.
• Urdu auto-switching: fixed an issue affecting automatic language switching when reading Urdu text.