Transcribing a file is one of the main features of Transcription App. Once a media file is uploaded to a project, you add it to the queue and process it with one of the speech recognition models. This tutorial covers the complete workflow from uploading to opening the finished transcript.
Supported media types
The upload picker accepts audio files, video files, and .mkv files.
Uploading a media file
- Open your project. If you do not have one yet, see 4. Projects and how to create one.
- In the Files panel, click « New Media ».
- Click the drop zone or drag a file into it.
- Confirm the selected file name.
- Enter a Display Name.
- Optionally add a Description.
- Choose the Language.
- Decide whether to enable Word-level timestamps. When enabled, the app tries to keep timing at the word level. This is required if you also want Speaker segmentation.
- If Word-level timestamps are enabled, optionally enable Speaker segmentation to split the transcript into speaker turns after transcription.
- Click « Upload ».
Tip: You can set a Default Language in Settings > Transcription so it pre-fills automatically. See 8. Guide to Settings.
Adding a file to the queue
On the media file card, click « Transcribe ». If the file already has a transcript, the app asks whether you want to replace it.
Processing the queue
- In the Queue panel, click « Click here to transcribe items ».
- If no model is loaded, the Load Whisper Model window opens. Select a model, choose the device and dtype, then click « Start Processing ».
- Wait for the file status to show Transcribed.
- Click the file card to open the transcript.
Use « Stop » if you need to halt processing immediately. Use « Clear Processed » after the run to remove completed and failed entries from the queue view.
Batch transcription
- Add multiple media files to a project.
- Click « Transcribe » on each file you want in the queue.
- Click « Click here to transcribe items ».
- Let the queue process all pending items.
- Use « Stop » if you need to halt processing immediately.
- Use « Clear Processed » after the run to remove completed and failed entries from the queue view.
Editing media metadata
On a media card, click « Edit » to reopen the file settings. You can change the display name, description, language, word-level timestamps, and speaker segmentation settings. Note that deleting a media file also deletes its transcript.
Choosing a model
In free mode, transcription uses Whisper Base. With a subscription, the best starting point is Whisper Large Turbo with dtype set to auto; if loading fails, try Whisper Small with auto.
- Whisper Base — fastest, lightweight. The free-mode model, good for rough drafts or weak hardware.
- Whisper Small — balanced speed and quality.
- Whisper Medium — higher accuracy, slower processing.
- Whisper Large Turbo — best balance of quality and speed. The main recommended option.
- Parakeet v3 (NVIDIA) — built for speed and well suited to long recordings. Covers 25 European languages; set the file’s language for the best results.
If model loading fails, the app shows a message explaining that the combination is not compatible with your computer. Try a smaller model, switch to auto, try a different dtype, or switch between CPU and GPU. For full model and dtype guidance, see 8. Guide to Settings.
What to expect for large files and video
- Video files are transcribed from extracted audio.
- Large media files may take longer to prepare before transcription begins.
- Very large files can spend noticeable time in preprocessing before text appears.
Free mode
Free mode is free forever: it transcribes complete recordings, with no duration limit, using Whisper Base. See 3. Getting a subscription to unlock Whisper Large Turbo, Parakeet v3, Whisper Medium and Whisper Small.