Support

We offer a 7-day free trial. During the trial period, you can try all features without any restrictions.

After the trial period, you can continue using the app by purchasing a license:

• Personal: ¥4,800 (per PC)
• Business: ¥59,800 (unlimited PCs)

Go to the license purchase page

Speech recognition works completely offline after the initial model download.

When using the AI processing feature:

• Cloud API (OpenAI / Anthropic / Groq / Gemini): Internet connection required
• Local LLM (LM Studio / Ollama): No Internet connection needed (completely offline)

Voice data is not sent anywhere. All speech recognition processing is completed entirely on your computer.

When using the AI processing feature, only the recognition result text is sent:

• Cloud API (OpenAI / Anthropic / Groq / Gemini): Recognition result text is sent to each provider's servers
• Local LLM (LM Studio / Ollama): No data is sent externally (completely offline)

For details, please see our Privacy Policy.

Recognition accuracy varies depending on the model used.

• large-v3-turbo: High accuracy & fast (Japanese recognition rate 93%+) [Default / Recommended]
• large-v3: Highest accuracy (Japanese recognition rate 95%+)
• medium: High accuracy (Japanese recognition rate 90%+)
• small: Good accuracy (Japanese recognition rate 85%+)

With clear audio, models of large-v3-turbo or above provide practical accuracy.

Yes, it works with CPU only. However, having a GPU significantly improves recognition speed. Without a GPU, we recommend using the large-v3-turbo or small model.

Currently, RocketWhisper is available for Windows only. Support for other platforms is being considered for the future.

Yes, it supports Japanese, English, Chinese, and Korean. The automatic language detection feature can also identify the spoken language automatically.

Yes, you can save results in multiple ways.

• Save as a text file (.txt)
• Save as SRT/VTT subtitle files
• Copy to clipboard
• Automatically saved to recognition history (viewable later)

Yes, long audio files can be processed. Files are automatically split into appropriate segments for processing. However, processing time depends on your PC's specifications.

Register the terms you want recognized in the "Custom Terms" tab in Settings. Using Whisper's initial prompt feature, registered terms will have improved recognition accuracy.

Additionally, the misrecognition correction rules in the "Text Processing" tab allow you to set up text replacement after recognition.

In the "App-specific Processing" tab in Settings, you can configure different processing modes for each application. For example, you can use polite language for email clients and casual tone for notepad.