RocketWhisper
Live Translation 🎬 Built-in ✨ v2.2.5
Windows version available →

Voice to Text. Instant. Accurate. Fully Offline.

Apple Intelligence powered. Blazing-fast Apple SpeechAnalyzer + on-device AI formatting.
No API keys. Fully offline. Your data never leaves your device.
New: real-time subtitle translation for audio from any video.

0
Speech Engines
0
LLM Providers
0
Offline Recognition
$0
Monthly Fee
RocketWhisper - AI voice input for Mac — Apple Intelligence, fully offline | Product Hunt
Apple Intelligence
Blazing-fast SpeechAnalyzer on macOS 26 + on-device Foundation Models for API-key-free AI formatting
Neural Engine Accelerated
Real-time recognition powered by Apple Silicon's Neural Engine and Metal GPU
🔒
Complete Privacy
Audio data never leaves your device
🎯
High Accuracy
7 punctuation rules and custom terminology dictionary for natural text output
Screenshots

See it in action

One click from your menu bar. Every feature on a clean, intuitive tab.

Model tab - speech engine selection
Model — Speech Engine Switch between Apple SpeechAnalyzer (macOS 26+) and four WhisperKit models with a single click.
Text Processing tab - punctuation, dictionary, corrections
Text Processing Auto punctuation, word dictionary, regex-powered correction rules. Seven-stage rule engine tuned for natural output.
Launcher tab - voice commands and voice launcher
Launcher — Speak to Act Voice commands, voice launcher, voice search. Say “Search for ...” and the right app or URL opens instantly.
Per-App tab - per-application processing modes
Per-App Modes Punctuation in your editor, casual in chat. Automatically switch processing per app you're focused on.
For MacBook Neo — A18 Pro · 8GB RAM · ¥99,800

AI voice input, from the moment
you open your MacBook Neo.

Apple’s new entry-level Mac — the MacBook Neo — is “Built for Apple Intelligence.”
RocketWhisper v2.2.5 is tuned to run instantly on the Neo’s 8GB RAM / 256GB SSD with zero model downloads and zero API keys.

0 MB
Model Download
SpeechAnalyzer ships with macOS 26
0
API Keys Needed
Foundation Models for AI formatting
$0
Cloud Cost
Everything runs on the Neural Engine
🔧 The Old Friction
Whisper models are heavy for an 8GB Neo
The entry-level Neo ships with tight storage and memory budgets. Downloading hundreds of megabytes to several gigabytes of Whisper models, then loading them into RAM on every launch, has never felt great on a machine like this.
  • 500 MB – 3 GB Whisper model download
  • Eats into the 256 GB SSD
  • Minutes of waiting on first launch
✨ Our v2.2.5 Answer
SpeechAnalyzer is the default on macOS 26+
Starting with v2.1.0, the default speech engine on macOS 26+ has been Apple’s built-in SpeechAnalyzer. The model already lives inside the OS, so RocketWhisper is ready the instant you install it.
  • No downloads — launches immediately
  • Neural Engine + unified memory = low overhead
  • Foundation Models AI formatting, no API keys
A “Built for Apple Intelligence” voice input
for a “Built for Apple Intelligence” Mac.
On M1–M4 Macs, you can still switch to WhisperKit anytime for maximum flexibility.
Why RocketWhisper?

What macOS Built-in Dictation Can't Do

RocketWhisper provides advanced features not available in macOS built-in dictation.

macOS Built-in Dictation
  • × No custom terminology dictionary
  • × No auto-correction rules for misrecognition
  • × No AI-powered text formatting
  • × Cannot edit selected text with AI
  • × Limited voice commands
  • × No per-app processing modes
  • × No punctuation control
  • × Cannot launch apps by voice
  • × No voice search
  • × Sends audio to Apple's servers
RocketWhisper PRO
  • Word Dictionary for company names, technical terms
  • Auto-correction rules with regex support
  • AI formatting with Apple Intelligence + GPT-4o / Claude / Gemini
  • AI Commands to edit selected text by voice
  • Voice commands like "new line", "delete"
  • App-specific modes with auto-switching
  • 7-stage auto punctuation engine
  • Voice Launcher for apps & URLs
  • Voice Search - "Search for..." triggers Google
  • Apple SpeechAnalyzer for blazing-fast recognition (macOS 26+)
  • 100% local processing, no data sent
Features

21 Premium Features

Uncompromising features designed for professional demands.

🎤

High-Accuracy Recognition

High-accuracy recognition powered by Neural Engine and CoreML. Choose from 4 WhisperKit models + Apple SpeechAnalyzer.

🔒

Fully Offline

All speech recognition is processed on-device. No internet required. Safely transcribe confidential information.

Global Shortcut

Default ⌥Space starts voice input from any app. Customizable. Right Option key also supported.

AI Text Formatting

Supports 6 providers: Apple Intelligence, OpenAI, Claude, Gemini, Groq, and Local LLM. Apple Intelligence requires no API key. Grammar correction, business style, summarization, translation auto-processed.

🔍

Voice Search

"Search for...", "What is...", "Look up..." - 10 voice command patterns instantly trigger Google search.

🚀

Voice Launcher

Launch apps or open URLs by voice. Say a keyword to instantly access your favorite tools.

💻

App-Specific Modes

Automatically apply different processing settings per app. Punctuation for editors, casual style for chat.

💬

Voice Commands

Edit text hands-free with voice commands like "new line", "paragraph", "delete". 7 built-in commands.

Auto Punctuation

7-stage punctuation rules optimized for natural text output, automatically inserting commas, periods, and question marks.

🛠

Auto-Correction Rules

Regex-supported correction rules for automatic misrecognition fixes. 27 hallucination filters built-in.

🎶

Floating Waveform Indicator NEW

Mini equalizer-style waveform bar during recording. Draggable, always-on-top for visual confirmation.

Right Option Hold Mode NEW

Record while holding Right Option, auto-stop on release. Push-to-Talk style for intuitive voice input.

🌐

Fn Key Push-to-Talk NEW

Push-to-Talk with Fn key (🌐). Double-tap to toggle continuous recording. Same feel as Wispr Flow or macOS Dictation.

📂

Batch Processing

Batch transcribe multiple audio files. Drag & drop to add, export as TXT, SRT, or VTT format.

Processing Pipeline

6-Stage Text Processing Pipeline

Intelligent processing flow that automatically transforms recognized text into high-quality output.

Stage 0
🚀 Launcher
Launch apps by keyword
Stage 0.5
🔍 Voice Search
"Search for..." triggers Google
Stage 1
💬 Voice Commands
Detect new line, delete, etc.
Stage 2
📚 Dictionary & Correction
Term replacement & fixes
Stage 3
✎ Punctuation
7 rules for natural punctuation
Stage 4
✨ AI Formatting
LLM polishes the text
AI Integration

Integrated with 5 AI Providers

Choose the optimal AI based on your use case, budget, and privacy requirements.

OpenAI
GPT-4o
GPT-4o mini
GPT-4 Turbo
Claude
Sonnet 4.5
Haiku 4.5
Opus 4.5
Gemini
2.5 Pro
2.5 Flash
2.0 Flash
Groq
LLaMA 3.3 70B
LLaMA 3.1 8B
Ultra-fast, free tier
Local LLM
LM Studio
Ollama
Fully private

Built-in Templates

💼 Business
🙌 Casual
📑 Summary
🌐 Translation
🔧 Grammar Fix
Custom
Whisper Models

4 AI Models

Select the optimal model based on speed and accuracy balance. All run on-device.

Model Size Speed Accuracy Use Case
Small 500 MB ⚡⚡⚡⚡ Real-time input
Medium 1.5 GB ⚡⚡⚡ Balanced
Large V3 Turbo Recommended 1.6 GB ⚡⚡⚡ High accuracy & speed
Large V3 3.0 GB ⚡⚡ Maximum accuracy

* For Japanese speech recognition, Large V3 Turbo or higher is recommended. Small/Medium may have reduced accuracy for kanji and katakana words.

Specifications

System Requirements

💻 System Requirements

  • macOS 14.0 Sonoma or later
  • Apple Silicon recommended (M1 / M2 / M3 / M4)
  • RAM 8GB or more (16GB recommended)
  • Storage 200MB + models (up to 3GB)
  • Microphone input

🎤 Input/Output

  • Input: Microphone (real-time recording)
  • Output: Direct text input / Clipboard
  • Shortcut (⌥Space) / Right Option / Fn key (tap & Push-to-Talk)
  • AI Commands (⌃⇧Space) for selected text editing
  • Auto-switching based on app detection

🌐 Supported Languages

  • Japanese (primary target)
  • English / Chinese / Korean
  • French / German / Spanish
  • Portuguese / Italian / Russian

🔐 Privacy

  • 100% local speech recognition
  • No external transmission of audio data
  • AI formatting uses your API keys directly
  • App Sandbox + Hardened Runtime
RocketWhisper

Accelerate your work with your voice.

7-day full trial. No credit card required.

v2.2.5 | macOS 14.0 Sonoma or later | Apple Silicon & Intel | Live Translation requires macOS 26+

Frequently Asked Questions

Is RocketWhisper free?+

All features are available for free during the trial period. After that, a one-time license purchase (Personal: ¥4,800 / ~$32) is required. No monthly subscription fees ever.

Does it work without internet?+

Yes, completely offline. All audio data is processed locally on your Mac and is never sent to any external server.

How accurate is the recognition?+

Powered by OpenAI Whisper’s latest model (large-v3-turbo), it delivers industry-leading accuracy. 73 hallucination countermeasures and custom vocabulary support ensure practical results.

Does it support Apple Silicon?+

Yes, runs natively on Apple Silicon (M1/M2/M3/M4) for optimal performance. Also works on Intel Macs.

Does it run well on the MacBook Neo (A18 Pro / 8GB RAM)?+

Yes, RocketWhisper is optimized for the MacBook Neo. Since v2.1.0, the default speech engine on macOS 26+ is Apple's SpeechAnalyzer (OS-built-in), so the WhisperKit model download (several hundred MB to GB) is no longer required. Even with 8GB RAM / 256GB SSD, the Neo can run AI voice input right after install — without eating disk space or API keys. AI text formatting via Foundation Models also runs entirely on-device; no cloud traffic.

What is Live Translation? 🎬+

Added in v2.2.1, Live Translation listens to any audio playing on your Mac (meetings, videos, podcasts, and more) through Apple SpeechAnalyzer, translates it with Apple Translation, and shows Netflix-style floating subtitles on top of your desktop. Zero cloud traffic, no API keys, fully on-device — across 10+ languages. Watch English talks with Japanese subtitles, follow Chinese news in English, and so on. Drag the overlay to reposition, and fine-tune the number of subtitle lines and source-text visibility. Requires macOS 26 or later.

Can I transcribe meetings?+

Yes, both real-time and from recorded audio/video files via batch processing. AI processing can automatically summarize and reformat the text.

Can I create video subtitles?+

Yes, transcribes video files (MP4, MKV, AVI, MOV, WebM) and exports subtitles in SRT and VTT formats.