Fully Offline AI Speech Recognition & Transcription — One-Time Purchase
You want voice input but don't want to send audio data to the cloud...
Other speech recognition software makes too many errors to be useful...
Industry-specific terminology and proper nouns aren't converted correctly...
Adding periods and commas manually is time-consuming...
Powered by a world-class AI speech recognition engine. Industry-leading accuracy for multiple languages.
All speech processing happens entirely on your PC. No data is ever sent externally.
Register industry jargon and proper nouns. Leverages Whisper's initial prompt feature for improved accuracy.
AI understands context and automatically inserts punctuation at natural positions. Perfect text effortlessly.
If your PC has an NVIDIA GPU, speech recognition speeds up several times to 10x or more.
No configuration needed — just launch the app and it automatically detects your GPU for optimal acceleration.
Two powerful new features make speech recognition even more versatile.
Select text, press the hotkey, and say "translate this," "summarize," or "make it formal." AI processes your text instantly.
Just say "Search for Tokyo Tower" and a Google search opens automatically in your default browser. Research has never been smoother.
100% Offline Compatible
AI processing works fully offline & free with local LLMs (LM Studio / Ollama)!
Also supports cloud LLMs: OpenAI, Gemini, Claude, Groq, and more.
Like a walkie-talkie — hold the Right Alt key to record, release to start recognition.
No extra steps. Intuitive Push-to-Talk makes voice input even more comfortable.
Hold Right Alt → Speak → Release → Auto recognition. Perfect for short inputs.
Quickly double-tap Right Alt → Continuous recording starts. Tap again to stop. Ideal for longer text.
Register your most-used AI tasks to dedicated hotkeys. Just speak, and translation, rephrasing, summarization and more are applied automatically.
No need to say "translate this" every time like AI Command mode. Register up to 20 custom instructions.
E.g.: Assign "Translate to English" to Ctrl+Shift+1 → press the hotkey → speak in Japanese → press the hotkey again → Whisper recognizes → LLM translates to English → result auto-output
🌐 Translate to English / 💼 Business Style / 📝 Summary / ✔️ Grammar Fix — ready to use out of the box. Add your own custom instructions too.
Supports Whisper small/medium/large-v3-turbo/large-v3 models. Choose the best model for your needs.
No internet connection required. Use it on airplanes, in offline environments — anytime, anywhere.
Customizable shortcuts let you start voice input instantly from any app. Right Alt hold mode also supported.
Transcribe multiple audio and video files at once. Process meeting recordings, interviews, and video files (MP4, MKV, etc.) in bulk.
Past results are automatically saved. Search and reuse any previous transcription instantly.
Say "new line," "new paragraph," or "delete" to format text with your voice alone.
Custom rules and regex-based auto-correction. 73 built-in hallucination countermeasures included.
Register industry terms and proper nouns to boost accuracy. Leverages Whisper's initial prompt feature.
Automatically apply different processing settings per app. Formal for email, casual for notes.
Supports OpenAI, Gemini, Claude, plus local LLMs (LM Studio/Ollama). AI processing even fully offline.
When minimized to system tray, a floating window with real-time waveform shows recording status during hotkey recording.
Say a keyword to auto-launch a registered application. Say "Notepad" and Notepad opens.
Select text, press the dedicated hotkey, and give voice instructions to have AI translate, summarize, or rephrase instantly.
Say "Search for ..." and a Google search opens automatically in your default browser. Supports 10+ phrase patterns.
Hold the Right Alt key to record, release to stop automatically. Double-tap to toggle continuous recording mode.
Register your most-used AI tasks to dedicated hotkeys. Translation, rephrasing, summarization — up to 20 instructions, one key each.
Launch the app and download your preferred Whisper model. Smaller models download in just a few minutes.
Press the record button and speak, or drag & drop an existing audio file.
Copy text to clipboard, save to file, or paste directly into your active application.
Keep your thoughts flowing without stopping to type. Turn ideas into text instantly.
Quickly transcribe meeting recordings. Batch processing handles multiple files at once.
Audio data is never sent externally. Safely use voice input even with confidential information.
| OS | Windows 11 / Windows 10 (version 1809 or later) * 64-bit only. 32-bit is not supported. |
|---|---|
| CPU | Required: x64 (64-bit) processor Recommended: Intel Core i5 / AMD Ryzen 5 or higher |
| Memory | Minimum: 4GB RAM Recommended: 8GB RAM or more (8GB+ recommended for large-v3-turbo model) |
| Disk Space | Application: Approx. 200MB Models: 75MB to 2.9GB (depending on selected model) |
| Runtime | .NET 8.0 Desktop Runtime * Included in Full edition. Lite edition prompts for installation on first launch. |
| Other | Microphone (for real-time recognition) Internet connection (for model download & cloud LLM AI processing) * Internet not required when using local LLMs (LM Studio/Ollama) |
| Model | File Size | Recommended RAM | Best For |
|---|---|---|---|
| medium | 1.5GB | 8GB | Short utterances under 5 seconds |
| large-v3-turbo | 1.6GB | 8GB | 5-20 second utterances (fast) |
| large-v3 | 2.9GB | 16GB | Long utterances over 20 seconds (highest accuracy) |
| Framework | .NET 8.0 WPF |
|---|---|
| Recognition Engine | OpenAI Whisper (Whisper.NET 1.9.0) |
| Supported Languages | Japanese / English / Chinese / Korean / Auto-detect |
| Supported Input Formats | WAV / MP3 / FLAC / OGG / M4A / WMA |
| Output Formats | Text / Clipboard copy / Direct paste |
| GPU Acceleration |
NVIDIA CUDA Supported (auto-detected) Automatically accelerated with NVIDIA GPU. Works fine on CPU without a GPU. |
🍎 Mac version | 🐧 Linux version
Latest Version: 1.2.0
Approx. 50MB
Recommended for users with fast internet. Downloads the AI model on first launch after installation.
Approx. 1.7GB
Ideal for fully offline environments. Includes the AI model and .NET Runtime — everything you need.
| Item | Lite Edition | Full Edition |
|---|---|---|
| Installer Size | Approx. 50MB | Approx. 1.7GB |
| AI Model | Downloaded on first launch | large-v3-turbo included |
| .NET Runtime | Separate installation required | Included (auto-installed) |
| Offline Use | Available after model download | Available immediately after install |
| First Launch | Usable after model download | Ready to use immediately |
| Best For | Fast internet environments | Offline & security-focused |
System Requirements: Windows 11 / 10 (64-bit), 8GB+ RAM recommended, .NET 8.0 Desktop Runtime
Yes, 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.
Yes, RocketWhisper works completely offline. All audio data is processed locally on your PC and is never sent to any external server. Perfect for privacy-conscious users.
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.
No, it works on CPU alone. If you have an NVIDIA GPU, CUDA acceleration is automatically enabled for significantly faster recognition. No setup required.
Yes, you can transcribe in real-time or from recorded audio/video files via batch processing. AI processing features can automatically summarize and reformat the text.
Yes, RocketWhisper can transcribe video files (MP4, MKV, AVI, MOV, WebM) and export subtitles in SRT and VTT formats for YouTube and other platforms.