Is internet access required for first-time use?

Internet is usually required during the initial installation and model download phase; once models are ready, the core transcription process runs entirely offline.

Which model should beginners download first?

It's recommended to start with light or extreme speed models to prioritize processing speed; switch to high precision models when higher accuracy is required.

Does the lifetime license limit device usage?

According to the documentation, authorization follows the Microsoft account and can be installed on multiple devices. Please refer to Store rules for current details.

Back Home

Getting Started

Owl Meeting is a 100% local-running Windows voice productivity tool. All speech recognition and AI processing are completed on your computer without the need for an internet connection, and data is never exported. Following this guide, you can complete your first transcription in minutes.

Last Updated: 2026-04-21 · Language: English

1. Installation and System Requirements

Install Source: Search for "Owl Meeting" in the Microsoft Store to install and enjoy automatic updates.
Runtime Dependency: The speech recognition engine requires Visual C++ Redistributable runtime support. If a prompt appears during the first launch, please download and install it as guided, then restart the software.
System Requirements: Windows 10/11, x64 architecture. No dedicated GPU is required; the CPU of an ordinary office computer is sufficient for smooth operation.

2. Model Download

After the first launch, you need to download AI models in the [Settings] → [Models] page. Once model download is complete, it can run fully offline:

Model management and download interface

Model	Function Description	Speed
Model 1 (Recommended)	Supports Chinese, English, Japanese, and Korean. Extremely fast, suitable for most scenarios. Preferred for new users.	Extreme Speed
Model 2	Good at Chinese dialect recognition. Needs to be used together with the punctuation model.	Normal
Model 3	Supports English and 26 European languages (Italian, Spanish, German, French, etc.).	Extreme Speed
Model 4	Widest coverage: Mandarin, Chinese dialects, English, Japanese, Korean, Russian, French, German, Arabic, etc., over 30 languages.	Normal
Punctuation Model	Supports Chinese and English punctuation completion. Can be enabled to fix punctuation issues in recognition results.

3. Quickly Start Your First Transcription

Owl Meeting provides two core working modes:

Real-time Transcription (Online Mode)

Suitable for ongoing meetings, lectures, or video calls:

Click [Online] to enter the real-time transcription interface.
Select sound source: Microphone (face-to-face meetings), System Sound (webcasts/videos), or Dual-channel mode (Tencent Meeting/Zoom, etc.).
Select recognition model; Model 1 is recommended for new users.
Click [Start Recording], and text will be displayed on the screen in real time.

For detailed parameters and advanced features, please refer to the Real-time Transcription documentation.

File Transcription (Offline Mode)

Suitable for processing existing audio or video files:

Click [Offline] to enter the file transcription interface.
Drag audio/video files into the window, or click [Select File]. Supports MP3, WAV, M4A, MP4, MKV, and other mainstream formats.
Select recognition model and segmentation method on the right. Model 1 + Time Interval Segmentation is recommended for new users.
Click [Start Recognition]; processing progress will be displayed in real time.

File transcription operation interface

For detailed parameters and advanced features, please refer to the File Transcription documentation.

4. Recommended Configuration for New Users

If you are a first-time user, the following configuration can help you get started quickly:

Recognition Model: Model 1 (Fast, good general-purpose)
Segmentation Method: Time Interval (VAD auto-segmentation, no extra config needed)
Recognition Mode: Normal
Denoising: Off for clear audio, on for noisy environments

Advance features such as speaker separation, Speaker Recognition, Custom Dictionary, and AI Assistant can be used later to gradually optimize effects.

Core Tip: All Owl Meeting transcription engines run locally; the processing does not rely on the internet at all. Your meeting recordings and transcripts always stay on your computer.