Smart Visionary 1.4.0

Smart Visionary is an NVDA add-on for AI-assisted dictation, clipboard rewriting, summarization, translation, screen and image description, OCR, PDF processing, media transcription, text-to-speech, result export, settings backup, and AI chat.

Menu-first operation

This version does not assign default keyboard shortcuts. Open NVDA Tools > Smart Visionary as a submenu to run add-on features directly. Users can still assign their own shortcuts from NVDA Input Gestures under the Smart Visionary category.

Stability and cleanup

Heavy AI, OCR, PDF, media, and TTS tasks run in the background and only one heavy task is allowed at a time to protect NVDA responsiveness. Temporary dictation recordings are removed after transcription succeeds or fails. The add-on also releases recording resources during NVDA shutdown or restart.

Prompts, styles, and dialects

Feature prompts now use strict output rules that ask AI systems to return only the final result without Markdown wrappers, introductions, or endings. Writing styles are managed separately from language and dialect settings, so a user can combine any supported style with any supported dialect. Custom writing styles can be created and edited from the settings dialog.

Text to speech

Text-to-speech has a dedicated dialog for engine, voice, language, style, dialect, emotion, lip-sync pacing, save mode, and output folder. Saved Gemini audio defaults to Downloads\SmartVisionary\Audio.

PDF and scanner tools

PDF text extraction first tries an available pure-Python PDF extractor, then falls back to Gemini OCR for scanned PDFs. Exported PDF text is saved as UTF-8 TXT to avoid unreadable symbols. Document tools include PDF to TXT, PDF to Word, PDF to audio, PDF to image, PDF merge, PDF split, PDF encryption, PDF decryption, TXT to PDF, Word to TXT, Word to image, Word to audio, image OCR, batch image OCR, image-to-PDF merge when Pillow is available, and black-and-white image conversion.

Translation and AI characters

A long-text translation command lets the user paste text, use the configured source and target languages, and translate with AI correction while preserving meaning and structure. The AI character/persona manager lets users create custom virtual characters with their own prompts.

No-key mode

Cloud providers require valid API keys. Local Ollama models are exposed as the no-key option and require Ollama to be running locally with the selected model installed.

Compatibility

The add-on targets NVDA 2019.1 and later, with fallbacks for older NVDA settings and menu APIs where possible.