Self-learning dictionary
Learns the proper nouns and jargon you actually use — and remembers the words you've corrected.
Spell a word wrong once, correct it, and it stays right from then on — local and cloud engines included.
VOICE → CONTEXT-AWARE WORK
From spoken thought to finished work.
VoiceWiz turns natural speech into clean, contextual text for coding, writing, research, and communication.
How it works
One pipeline.Configured around your work.
Choose how VoiceWiz transcribes, what context it uses, and whether it refines the result — all from one configurable workflow.
voicewiz / transcribe
1 / 3
01 / Transcribe
awaiting inputPick an engine
Choose the transcription provider and model that fit this recording.
Transcription source
Transcription modelselect a provider
Select a transcription source to choose a model.
Batch transcriptionselect source
02 / Context
awaiting inputGive it the right frame
Choose a mode and VoiceWiz keeps the relevant working context close to your words.
Pick a mode above to load its context.
03 / Refine
optionalShape the result
Turn the transcript into something ready to use, with a model that fits your workflow.
Prompt Builder
Apply the selected model to polish structure and tone.
Refinement provider
Refinement modelselect a provider
Select a provider to choose a refinement model.
Lenient LLM timeout: autooptional
Choose the transcription engine that fits your setup.
Use cases
One voice. Different kinds of work.
Tell VoiceWiz what kind of work you're doing and let the pipeline shape the result.
You say
- He knows going home is a bad idea.
- Do not say it directly.
- Keep my usual restrained style.
VoiceWiz returns
DIRECTION
Let his practical reasoning conceal the fact that he does not actually want to return.
SUBTEXT
Show hesitation through what he notices, delays, or avoids rather than direct explanation.
STYLE
Keep the prose restrained, close to the existing chapter voice, with sparse internal commentary.
illustrative example — not a captured session
Capabilities
It gets smarter every time you use it.
Beyond the pipeline — the more you use VoiceWiz, the more it knows about how you work.
Example mining
Feed it papers, manuscripts, or code commits. It mines style patterns and retrieves the right ones for each task.
Your files become a searchable style library — classified by writing domain and pinned per project.
Working-file continuity
Point it at the file you're editing. It reads the live tail so every dictation continues your draft, not a blank page.
Follows the live buffer even before you save, and jumps to the current section on its own.
Semantic note search
Search by meaning, not just keywords — vector and full-text search fused, with a keyword fallback if the index is ever down.
Merges semantic and keyword hits into one ranked list — still fully offline.
Citation enforcement
When you cite a source, VoiceWiz checks it against the sources you've linked and quietly fixes or removes anything that doesn't check out.
Produces ready-to-paste references in APA, MLA, Chicago, IEEE, and Harvard.
Prompt Studio
Drop an MP3, Markdown file, or PDF in and run it through the same pipeline — no recording needed.
Transcribes audio in batches with live progress, then streams the finished text into the pipeline.
Why it matters
Skip the AI detour.
A normal AI workflow moves a thought through a transcription tool, the clipboard, an AI provider, and back to your work. VoiceWiz keeps the processing close so you can speak and paste at the cursor.
Workflow comparison: the normal path is Speak, Copy transcription, Paste into AI provider, Insert transcription and prompt, AI provider gives output, Copy output, and Paste at desired location. With the VoiceWiz workflow, the path is Speak, VoiceWiz processes transcription, and Paste at cursor.
FAQ
Questions, answered.
VoiceWiz is a desktop voice-to-work application. It turns spoken language into clean, contextual text — for programming, writing, research, and communication — instead of just returning a verbatim transcript.
Transcription is only the first stage. VoiceWiz adds the context of what you're working on and an optional refinement stage, so the output is shaped for the task instead of needing cleanup. If you want plain dictation, Direct mode provides exactly that.
Programming, writing, research, emails, notes — any text-heavy professional workflow where speaking is faster than typing but raw speech isn't yet usable text.
Yes. Local transcription models (whisper-class models and NVIDIA Parakeet) and local refinement models run on your own machine — no API key required for the local pipeline. Supported cloud providers are available too if you configure your own keys.
VoiceWiz runs on Windows and Linux today. It will be free to access during the first month after release; pricing after the launch period hasn't been announced yet. See the download page for current builds, or the docs to learn more.
Your keyboard can take a break.
Speak the thought. Let VoiceWiz turn it into something you can use.
Free access during the first month after release.