How to transcribe audio to text offline on Windows
To transcribe audio to text offline on Windows, use an app that runs a speech recognition model on your own PC instead of uploading the file to a website. WizWhisp does this with OpenAI's Whisper models. It is free to download from the Microsoft Store, works without an internet connection, and keeps your recordings on your computer.
Why transcribe offline
- Privacy. The audio never leaves your PC, which matters for interviews, lectures, meetings, and anything confidential.
- No per-minute fees. Many online services charge by the minute or by subscription. An offline app costs the same however much you transcribe.
- No uploads. Long recordings start right away, with no file size limits or waiting for an upload.
The trade-off is that speed depends on your computer.
What you need
- A PC running Windows 10 or later, with at least 8 GB of RAM. The Microsoft Store listing shows the exact supported Windows versions.
- Optional: an NVIDIA or AMD graphics card. With an NVIDIA GPU, transcription can be up to 10× faster than on the CPU alone. Without a GPU, WizWhisp runs on the CPU.
- Your recording, for example an MP3, WAV, M4A, MP4, or AVI file.
- WizWhisp, free from the Microsoft Store.
Step by step
- Install WizWhisp from the Microsoft Store.
- Choose a model. Open WizWhisp and pick a model. Tiny and Small are free without limits. Large v3 turbo is the most accurate; it comes with Pro, and the free version lets you try it on 3 minutes of audio.
- Choose the language, or leave automatic detection on. WizWhisp supports more than 90 languages.
- Add your file. Drag your audio or video file into the window. Transcription starts right away; processing time depends on the length of the file, the model, and your hardware.
- Export the result. Save TXT for documents and notes, SRT or VTT for video captions, or LRC for timed lyrics.
Which model should you use?
| Model | Best for | Availability |
|---|---|---|
| Tiny | Quick drafts and slower PCs | Free |
| Small | Everyday recordings with clear speech | Free |
| Large v3 turbo | Difficult audio: noise, accents, technical terms | Pro; 3-minute trial in the free version |
Start with Small. If the result needs many corrections, try Large v3 turbo on a short section to compare.
Tips for better accuracy
Transcribe audio that plays on your PC
To transcribe an online class, a meeting, or a stream, record it first and then transcribe the recording. NowSmart Audio Recorder records any sound your PC plays; see How to record computer audio on Windows 10/11. Ask for consent before you record other people.
FAQ
Is WizWhisp free?
Yes. WizWhisp is free to download, and the Tiny and Small models work without limits. Pro is a one-time USD 9.99 purchase with no subscription. It adds the Large v3 turbo model, a batch queue for many files, and support for your own compatible Whisper models.
Does WizWhisp upload my audio?
No. Transcription runs entirely on your PC. WizWhisp does not upload files, collect data, or show ads.
Does it work without an internet connection?
Yes. WizWhisp works fully offline.
Does WizWhisp work with AMD graphics cards?
Yes. WizWhisp supports NVIDIA and AMD GPUs, and it also runs on the CPU.
How long can a recording be?
There is no length limit. WizWhisp decodes the audio as it transcribes instead of loading the whole recording into memory, so multi-hour interviews and lectures do not use much extra memory. Longer files take longer to process.
Which languages are supported?
More than 90 languages, with automatic language detection or a manual choice.
Can WizWhisp tell different speakers apart?
Not currently. Offline speaker identification is not yet reliable enough; it may be added later.
Start with the free models
WizWhisp's Tiny and Small models are free without limits. WizWhisp Pro adds Large v3 turbo, a batch queue, and custom models for a one-time USD 9.99.