Working with VoiceWiz

Around the core pipeline sit the tools that make dictations reusable: History for finding past work, the Library for style and reference material, the Dictionary for your vocabulary, and Studio for running existing text through the pipeline.

History

Every dictation is saved locally — raw transcript and refined output together — so nothing you've spoken is lost. History lives in the Control Panel and is stored in a local database; it is not synced anywhere.

• Search — full-text search across all past transcriptions.
• Semantic search — search by meaning, powered by a local embedding model and vector index. If the vector engine is unavailable, keyword search still works.
• Manage — copy any past result, edit or delete entries, or export a transcription as a note.

Library

The Library keeps the raw material for better refinement: example texts that teach the Prompt Builder how you like your output to read, and your own documents, mined for style and content.

Example stacks

Examples are short input/output pairs that demonstrate the transformation you want. Grouped into stacks, they're retrieved when relevant and injected into the prompt as reference patterns.

Mining documents

Feed VoiceWiz your own writing — papers, chapters, reference material (PDF, DOCX, Markdown, plain text) — and it extracts style samples that match your writing voice. For research work, documents can be indexed by full content into a local vector store; on each research dictation, the most relevant passages from your literature are retrieved as grounded context.

Dictionary

Speech models are trained on generic language — they don't know your product's codename, your colleague's surname, or the library you work with. Words you add to the custom dictionary are passed to the transcription backend as hints, with both local and cloud backends.

1. 1Open the Control Panel and go to Dictionary.
2. 2Type the word or phrase — add several at once, separated by commas.
3. 3Press enter. The words are active immediately for future dictations.

The dictionary can also grow on its own:

• Learn from my dictation history — names and terms that appear repeatedly are added silently.
• Auto-learn from corrections — when you correct a mis-transcribed word in the target app, the corrected spelling is added automatically.
• Built-in spelling corrections — fixes common speech-to-text errors for your language.

Studio

Studio is the pipeline without the microphone: paste text you already have, and run it through the same context-and-refinement stages your dictations use. Useful for reworking typed text, transforming when speaking isn't an option, or iterating with a second pass. Studio results are saved to History together with the mode and context used.