Installation Guide

Set up VoicePrompt™ for Windows when the signed download becomes available.

System Requirements

• Windows 10 or Windows 11 (64-bit)
• 4GB RAM minimum (8GB recommended for local Whisper models)
• 2GB free disk space (for Whisper model downloads)
• Internet connection for API calls and model downloads (optional for offline transcription)
• Microphone (built-in or external)

Download & Install

1. The Windows download and installation instructions will be published after the signed release is verified.
2. On first run, open Settings and choose a speech model and microphone.
3. VoicePrompt keeps a tray icon available for Settings and status.

System Tray Menu

Right-click the VoicePrompt icon in the system tray to access:

• Settings: Open the main settings window
• Status and active mode: See whether dictation is ready and which mode is active
• Retry last dictation / Copy last transcription: Recover a take when those actions are available
• Exit: Close VoicePrompt™ completely

Initial Setup

1. Get Ready to Dictate

First run walks through three checks in one window: check live microphone input, download or choose a local Whisper model, and press the actual recording shortcut. Setup finishes when all three are ready. A smaller model downloads faster; start with the recommended model unless you need a smaller download or more accuracy.

Basic local dictation needs no provider key or paid license. You can defer setup and return through Settings. Optional provider configuration is available in Settings → API & Models.

2. AI Provider (Optional)

Basic dictation can run offline with a local model and LLM Model set to None. Cloud transcription and AI cleanup need a supported provider connection; most providers use your own API key.

Supported providers:

• OpenAI (GPT family)
• Anthropic (Claude)
• Google (Gemini API)
• Groq (Llama via OpenAI-compatible endpoint)
• Custom OpenAI-compatible endpoint

Add keys in Settings → API & Models. Google sign-in is available only in builds where it has been enabled; public availability is pending Google's production verification.

3. Choose Your Hotkey

New installs default to Ctrl + Alt + D.

To change it:

1. Right-click the system tray icon
2. Select "Settings"
3. Go to "Hotkeys" tab
4. Click "Record" and press your desired combination
5. Use the shortcut check and press the combination when prompted. VoicePrompt confirms an actual Windows hotkey event; that check does not record audio.

Provider Setup (Optional)

Configure a provider in Settings → API & Models for optional cloud transcription, AI cleanup, and Assistant mode. Provider prices and limits vary.

OpenAI

1. Create a key at OpenAI
2. Paste into Settings → API & Models → OpenAI
3. Click Test

Anthropic (Claude)

1. Create a key at Anthropic
2. Paste into Settings → API & Models → Anthropic
3. Click Test

Google (Gemini)

Public Google sign-in is pending production verification. If your build offers Sign in with Google, it authorizes VoicePrompt directly with Google and requires the relevant account and project access. It does not reuse a Gemini CLI login. You can also connect Gemini with your own API key from Google AI Studio in Settings → API & Models.

Groq (Llama)

1. Create a key at Groq
2. Paste into Settings → API & Models → Groq
3. Click Test

Custom Endpoint

Use an OpenAI-compatible endpoint by selecting Custom and entering the base URL and model name. Add a key if the endpoint requires one.

Using VoicePrompt™

Basic Usage

1. Click in any text field where you want to type
2. Press the dictation hotkey once to start (default: Ctrl + Alt + D)
3. Speak naturally and clearly
4. Press the hotkey again to stop
5. The overlay shows elapsed recording and processing time, then the delivery or recovery result. Text is inserted into the field you were using. If insertion fails, copy the last transcription from the tray or review its text in Settings → Recent dictations.

Offline Transcription Setup

VoicePrompt™ uses local Whisper models for offline transcription:

1. Open Settings → API & Models
2. Choose one of the available local transcription models
3. Use Download Model if the chosen model is not installed
4. Choose None under LLM Model to dictate without AI cleanup or a provider key

Note: Larger models need significantly more RAM. Choose a size that fits your PC’s memory to avoid slowdowns or load failures. Smaller models start faster; larger models can be more accurate on tough audio.

Tips for Best Results

• Speak at a normal pace - not too fast or slow
• Pause briefly between sentences for better accuracy
• Say punctuation naturally: "period", "comma", "question mark"
• Review names and technical terms, especially on the first take

Voice Commands

Spoken editing commands are available with a Personal or Professional license. Free dictation inserts the transcript as spoken.

Say a supported command as the entire recording. Commands are deliberately not triggered by longer sentences: “copy the report” is dictation, while “copy” on its own requests the copy action.

• Editing: undo, redo, copy, cut, paste, delete, select all.
• Keys and formatting: escape, enter, new line.
• Punctuation: period, comma, question mark, exclamation mark, colon, semicolon, quotation marks, apostrophe, dash, hyphen.

Say “undo” or “redo” at the end of a take to reverse or repeat the last supported action. For an AI rewrite such as “make this more formal,” use Assistant mode with AI processing and a configured provider.

Hotkey Configuration

VoicePrompt™ uses global hotkeys that work anywhere on your system, even when the app is minimized.

Default Hotkeys

• Main Recording Hotkey: Ctrl + Alt + D - Start/stop dictation
• Paste last take: Ctrl + Alt + V
• Paste original transcript: Ctrl + Shift + Alt + V

Changing Hotkeys

1. Open Settings (right-click system tray icon → Settings)
2. Go to the "Hotkeys" tab
3. Click "Record" beside the main recording hotkey
4. Press your desired key combination
5. The new hotkey will be saved automatically. Use the shortcut check and press it to confirm Windows delivers the event.

Global vs Assistant Mode

VoicePrompt™ supports two different modes:

• Global Mode (Main Hotkey): Transcribes your speech. Optional AI processing can clean up the result when configured.
• Assistant Mode: Sends your request to a configured AI model and inserts its response.

Context Modes

Context modes provide instructions for optional AI processing. Choose a preset in Settings → Context, or create one for your work. You can also enable automatic switching based on the focused app and each mode's app list.

Global is the default context. Other presets can guide style for email, coding, documentation, chat, or meetings. Results depend on your configured model and the words you dictate; review important output before using it.

Context modes do not turn a long dictation into a voice command. Privacy Mode disables online AI processing.

Voice Corrections

Basic dictation inserts the transcription. If you enable AI Processing and configure an AI model, VoicePrompt can clean up wording and apply the selected context mode. The result may differ from what you said, so review it before sending or publishing.

Use Assistant mode for a deliberate rewrite or question. Spoken editing commands such as “undo” and “redo” work as standalone recordings; see Voice Commands.

Screen and focused-app context are optional Settings → Context controls. A screenshot can be sent to your provider when screen context is enabled.

API Configuration

Local transcription needs no provider account. Cloud transcription, AI cleanup, Assistant mode, and some Computer Use routes need a supported provider. In Settings → API & Models, choose a provider and model, then add its key or use the available Google sign-in option.

A custom OpenAI-compatible endpoint needs its base URL and model name. It may also require a key. Provider availability, limits, and charges are controlled by that provider.

New provider keys are protected locally with Windows DPAPI. Google sign-in uses your own Google account and stores the authorization on this PC. Privacy Mode disables online processing.

See Provider Setup for provider-specific steps.

Recent Dictations

Open Settings → Recent dictations to review up to 100 recent text takes saved on your PC. Select a take to copy its output or original transcript, edit the saved output, or delete it. Editing a saved take does not change text already inserted into another app; copy and paste the revised text where you want it.

Remember dictations controls future saves. Turning it off keeps existing records until you delete them or use Clear all. Clear all asks for confirmation. This history stores text and context labels, not audio; Retry last dictation uses a separate temporarily retained recording.

Troubleshooting

No audio detected

Check the selected microphone in Settings → Audio, Windows microphone permissions, mute state, and the test recording in Settings → Voice & Transcription.

Provider request failed

Check your connection, selected provider and model, and that the key or Google sign-in is authorized for that service. For a custom endpoint, check its URL and model name. Provider limits and billing are managed by the provider.

Hotkey does nothing

Use the shortcut check in Settings → Hotkeys and press the shown combination. It reports whether Windows delivered the press, or whether the check timed out or registration failed. If another application owns the shortcut, record a different combination. During normal use, press once to start and once more to stop.

Text does not appear

Focus an editable text field before recording. Some apps block simulated input; use the tray's Copy last transcription command and paste it manually. An elevated target app may also reject input from an unelevated VoicePrompt.

Local model will not load

Check available disk space and memory, then try another model listed in Settings → API & Models. Restart VoicePrompt after a failed download.

Context does not switch

Check Settings → Context: automatic switching must be enabled, the context preset must be enabled, and its app list must match the focused application.

Tips for Reliable Dictation

• Choose the microphone you intend to use and make a test recording in Settings.
• Focus the text field, press the recording hotkey, speak, then press the hotkey again to finish.
• Keep background noise low. Shorter takes are easier to review and retry.
• Say editing commands as a separate take. “Copy the report” remains dictation; “copy” alone requests the copy action.
• Use local transcription and Privacy Mode when content must stay on your PC.
• Review AI-cleaned or Assistant output before sending it. For cloud features, check your provider's usage and charges in its own account.

Settings: General

Control privacy, startup, notifications, and the recording indicator.

• Privacy Mode: Keep dictation local and disable online AI features while it is on.
• Start with Windows / Minimize to System Tray: Choose how VoicePrompt starts and stays available.
• Show Notifications / Dictation Sounds: Control status messages and recording chimes.
• Enable Logo Animations: Animate the tray icon and its recording state.
• Show recording overlay: Show a compact Listening / Processing pill with elapsed time and the final delivery or recovery result. It does not take focus and is hidden in exclusive fullscreen apps.
• Updates: See the current version and distribution channel, check manually, or enable update notifications. Store installations use Microsoft Store; direct installations use the official release page. Background checks wait for completed setup and idle dictation. Updates are never installed automatically by VoicePrompt.

Settings: API & Models

Choose speech and AI models. Local dictation works without a provider key when LLM Model is set to None.

• Provider and API Key: Add a key for OpenAI, Anthropic, Google, Groq, or a supported custom endpoint. Use Test API Key to check the selected provider.
• Gemini: Sign in with Google using VoicePrompt's OAuth client, or supply a Gemini API key. Google account and project access affect sign-in availability.
• Transcription Model: Choose an available local or cloud speech model. Download a local model from this page if it is missing.
• LLM Model: Choose a model for AI cleanup and Assistant mode, or None for basic dictation without AI processing.
• Custom Endpoint: Enter the compatible endpoint URL and exact model name. A key is optional only when your endpoint does not require one.
• Open models folder: Find downloaded local models on your PC.

New provider keys are protected with Windows DPAPI on this device. Provider charges and limits are separate from a VoicePrompt license.

Settings: Voice & Transcription

Check the microphone signal and choose the transcription language.

• Audio monitoring: Watch the input level and use Test Recording to hear a short sample.
• Live Transcription Output: See recognized speech while the app is working.
• Language: Choose a language or Auto-Detect.
• Spoken undo and redo: Say either at the end of a take. The current page does not expose a command-list editor.

Settings: Audio

Choose the microphone under Input Device. The list refreshes when audio hardware changes.

Recording uses the dictation hotkey as a toggle: press once to start and again to stop. Hold-to-record Push-to-Talk is not available in this release.

Settings: Hotkeys

The main dictation shortcut starts and stops a take. New installs default to Ctrl + Alt + D.

• Choose Record beside Main Recording Hotkey, then press the combination you want. Use the shortcut check to confirm a real press reaches VoicePrompt; the check does not start a recording.
• Paste the last take and Paste original transcript show their assigned shortcuts on this page.
• Press Escape twice within 1.5 seconds to cancel a take in progress.
• Assistant mode has no separate global shortcut. Choose it in the app's settings or available mode controls.

If a shortcut collides with another app, choose a different combination. Ctrl + Shift + V conflicts with paste without formatting in Chrome and Edge.

Settings: Advanced

Control optional AI processing and Computer Use.

• Enable AI Processing: Allow cloud cleanup, Assistant mode, and other AI text improvements when a supported model is configured. Privacy Mode disables it.
• Enable computer use: Turn on paid desktop control only when you intend to use it. It is off by default.
• TypeSafe API key and model: Optional fast decision model for selecting labeled Windows controls. Leaving the model name empty uses jev-latest.

Computer Use

Optional voice-driven desktop control for Personal and Professional licenses. Free dictation does not include it.

1. Activate a paid license in Settings → License.
2. In Settings → Advanced, enable Computer Use. Configure either the optional TypeSafe key for fast labeled-control decisions or an OpenAI or Anthropic key in API & Models for screenshot-based control.
3. Use Assistant mode and say what you want to do. VoicePrompt can choose a control, click, type, and scroll in a Windows app.
4. Watch the session. Stop it with the recording hotkey if it is doing the wrong thing.

The TypeSafe route reads Windows UI Automation labels. The OpenAI and Anthropic route can capture screenshots. Your instruction and the relevant labels or screenshots may be sent directly from your PC to the configured provider. Provider fees and limits are separate. Privacy Mode disables online Computer Use.

Settings: Context

Choose how a configured AI model uses the active app and optional context.

• Include Screen Context: Capture a screenshot for AI context when enabled. Screen content may be sent to your configured provider.
• Include Focused App Context: Include information about the active app.
• Context Mode Auto-Switch: Change the active context based on the app and its configured app list.
• Presets and custom contexts: Enable presets, edit their app lists, create custom contexts, or pin a favorite with Always Use.
• LLM Replacements: Add phrases and replacement values for model-assisted writing.

Context presets and replacements need AI Processing and an AI model. Privacy Mode disables the online context features.

Settings: License

Manage your VoicePrompt™ license and view application information.

• License Activation: Paste the key from your purchase email and click Activate License. Activation needs internet access. You can refresh license status or deactivate an active license here.
• License Status: Shows current state (Unlicensed/Personal/Professional) with machine ID display.
• Application Information: Product Name (VoicePrompt™), Version, Build Date, Installation Date, Machine ID.
• Purchase Options:
  • Non-Commercial License: $29.99 launch price (normally $59.99, 50% off), for personal use on one device. Includes optional Computer Use.
  • Professional License: $149.99 launch price (normally $249.99, 40% off), for commercial use on two devices. Includes optional Computer Use.
• Availability: The launch offer is displayed now; checkout will open when licensing is ready. No countdown or end date has been announced.
• Support & Resources: Website, Documentation, Support, Contact Support, and Send Feedback links or buttons.
• Legal Information: End User License Agreement, Privacy Policy, Third-Party Licenses links.

Note: License activation is disabled when Privacy Mode is active.

Settings: Usage

Track your estimated VoicePrompt™ usage history and activity.

• Estimated Usage Summary: Four cards showing Estimated Calls, Estimated Tokens, Estimated Cost, and Transcription Minutes over selected time range.
• Controls:
  • Time Range: Last Hour, Last 24 Hours, Last Week, Last Month, Last Year, All Time
  • Refresh Data: Reload usage statistics
  • Export Data: Export usage data to file
  • Reset Statistics: Clear all usage data
• Usage Charts: Four charts showing Daily Cost, Daily Calls, Daily Tokens, and Transcription Minutes with visual indicators and last activity timestamp.
• Privacy: Usage data is stored locally only.