User Guide

1Installation

System Requirements

Installation Steps

1. Download the latest .dmg file from the download page.
2. Double-click the downloaded DMG file and drag the RocketWhisper icon to the Applications folder.
3. If Gatekeeper shows a warning on first launch, click "Open". Alternatively, go to System Settings > Privacy & Security and click "Open Anyway".
4. On first launch, the Whisper model will begin downloading. This may take a few minutes depending on your network.

Choosing a Model

2Basic Usage

Voice Recognition from Microphone

1. Click the RocketWhisper icon in the menu bar to show the popup window.
2. Click the Record button (microphone icon).
3. Speak into your microphone.
4. Click the Stop button.
5. The recognition result will appear in the text area.

Using Recognition Results

• Copy: Click the copy button to copy the result to your clipboard.
• Auto Copy: When enabled in settings, results are automatically copied on completion.
• Auto Paste: When enabled, results are automatically pasted into the active app's text field when using shortcut recording.

3Settings

Click the gear icon in the popup window to open settings. Various options are available in the following tabs.

4Global Shortcut

RocketWhisper supports customizable global shortcuts, allowing instant recording from any app. It also supports Right Option key tap and hold (Push-to-Talk).

Right Option Key Operations

Recommended Shortcut Settings

Cancel Recording

Press Escape during recording to cancel without processing.

5Recognition History

RocketWhisper automatically saves past recognition results. Click the History button in the popup to view the history list.

History Features

• List View: View past results with timestamps
• Search: Search past results by keyword
• Copy: Copy any history item to clipboard
• Delete: Delete individual history items
• Export: Export history as a text file

6Text Processing (Punctuation & Line Breaks)

RocketWhisper includes advanced text processing features to produce natural, well-formatted text.

Punctuation Prompt

A prompt is set for the Whisper model to encourage punctuation output, making the model more likely to include punctuation.

Auto Punctuation

Post-processing applies 7 rule-based punctuation insertions. Even if Whisper's output lacks punctuation, natural punctuation is added.

Auto Line Breaks

Automatically inserts line breaks at sentence boundaries, making long text more readable with paragraphs.

7Voice Commands

Voice commands let you execute text editing operations by speaking specific phrases. Enable/disable in "Text Processing" settings.

Supported Commands

If Voice Commands Aren't Recognized

Whisper may sometimes misrecognize homophones based on context.

If voice commands don't work as expected:

1. Check that voice commands are enabled — Verify "Voice Commands" is ON in the "Text Processing" settings tab.
2. Pause before and after commands — A short pause before and after helps distinguish commands from regular speech.
3. Speak clearly — Pronouncing each syllable distinctly improves recognition.

8Word Dictionary (Custom Terms)

The word dictionary allows you to register technical terms, company names, personal names, and acronyms that Whisper might not recognize well. This dramatically improves recognition accuracy. Not available in macOS built-in dictation.

How It Works

Registered words are used as WhisperKit promptTokens, making the Whisper model prioritize these terms in its output.

How to Register

1. Open the "Word Dictionary" tab in settings.
2. Click the "Add" button.
3. Enter the word to register (e.g., React, TypeScript, AWS).
4. Optionally set a reading/pronunciation (for Whisper recognition assistance).

Features

• Fully local processing — No API costs, no internet required
• Real-time effect — Active immediately after registration
• Dictionary replacement — Can also be used for auto-replacement rules

9Auto-Correction Rules

Set up rules to automatically correct misrecognitions in Whisper output. Supports both simple string replacement and regex.

Rule Types

• Simple replacement: Replace a specific string with another
• Regex: Advanced replacement using regular expression patterns
• Case sensitivity: Set per-rule whether to match case

Built-in Hallucination Filters

27 filters are built in to automatically remove "hallucination text" that Whisper sometimes generates during silence, such as:

• "Thank you for watching"
• "Please subscribe"
• "Good night" (hallucination during silence)

Preset Rules

Preset rules for common misrecognition patterns are available. Enable with one click from settings.

Custom Rules

1. Open the "Correction Rules" tab in settings.
2. Click "Add Rule".
3. Enter the search string (misrecognized text) and replacement string (correct text).
4. Optionally enable "Use regex" and "Ignore case" options.

10App-Specific Processing Modes

App-specific modes automatically apply different text processing settings based on the focused app. For example, use punctuation for text editors and casual style for chat apps.

Processing Modes

Setting Up App Mapping

1. Open the "App-Specific" tab in settings.
2. Enable "App-Specific Processing Modes".
3. Click "Add" to select an app and set its mode.

11AI Processing (LLM Integration)

RocketWhisper integrates with 5 AI providers for advanced processing like text formatting, translation, and summarization.

Supported AI Providers

Setup

1. Open the "AI Processing" tab in settings.
2. Select your provider.
3. Enter your API key (not needed for Local LLM).
4. Select the model to use.
5. Enable AI processing and choose a processing mode.

Local LLM Setup Examples

12AI Command Mode

AI Command mode lets you give voice instructions to edit selected text with AI. Perform translation, summarization, tone changes, and more with just your voice.

How to Use

1. Select text in any app.
2. Press ⌃⇧Space (Control + Shift + Space) to activate AI Command mode.
3. Speak your instruction (e.g., "Translate to English").
4. AI processes the selected text according to your instruction and replaces it.

Examples

13Custom Instructions

Custom Instructions let you pre-assign AI processing prompts to dedicated shortcuts. Unlike AI Commands, you don't need to speak instructions each time — the recognized speech is automatically processed using a pre-configured prompt.

Difference from AI Commands

How to Use

1. Create an instruction in the "Custom Instructions" settings tab and assign a shortcut.
2. In any app, press the assigned shortcut to start recording.
3. Speak into the microphone (the recognized text becomes the AI input).
4. Press the same shortcut again to stop recording.
5. Speech is recognized and processed by AI with the pre-configured prompt, then automatically pasted.

Preset Instructions

Four presets are automatically created on first launch. These can be edited but not deleted.

14Voice Launcher

Voice Launcher lets you launch apps or open URLs by speaking registered keywords. Processed at Stage 0 of the pipeline, so matching keywords execute actions without text output.

How It Works

• Exact match for keywords (ignoring punctuation and case)
• On match, launches the registered app or opens URL in browser
• No text output on match (action only)

Setup

1. Open Voice Launcher settings.
2. Click "Add".
3. Enter the trigger keyword (e.g., "Notes", "Browser").
4. Enter the app path or URL to open.

Configuration Examples

15Voice Search

Voice Search lets you trigger Google searches instantly by speaking specific phrases. Results open in your default browser.

Supported Phrases (10 Patterns)

16Menu Bar Icon

RocketWhisper is a menu bar resident app. It doesn't appear in the Dock. All operations are accessible from the menu bar icon.

Basic Operations

• Click: Click the menu bar icon to show the popup window.
• Recording status: The icon changes during recording to show current state.
• Auto-start: Add to login items to auto-start in the menu bar on Mac startup.

Setting Up Auto-Start

1. Open System Settings.
2. Go to General > Login Items.
3. Click "+" to add RocketWhisper.

17Floating Waveform Indicator

A floating window showing a small mini equalizer-style waveform during recording. Always on top, so you can confirm recording status even while working in other apps.

Specifications

Operations

• Drag to move: Drag the indicator to any position.
• Position memory: Position is saved and restored on next launch.
• All Spaces: Visible on all macOS desktops (Spaces).
• Always on top: Stays above all other windows.

Settings

Toggle "Show floating waveform during recording" in the "Model & Language" settings tab. Default is enabled (ON).

18Batch Processing

Batch transcribe multiple audio files at once. Convenient for processing recorded meeting audio or interview files.

How to Open

1. Open the menu bar popup.
2. Click the Batch Processing button (document icon) in the header.
3. A separate batch processing window opens.

Usage

1. Add files: Click "Add Files" to select audio files, or drag & drop onto the window.
2. Start batch: Click "Start Batch" to transcribe files sequentially.
3. View results: Recognition results (character count) are shown in the list.
4. Export: Select export format from the "Export" menu and choose a destination folder.

Supported File Formats

Export Formats

19Troubleshooting

If you encounter issues, refer to this FAQ.