Automatic transcription of your videos and audio
Polimake Desktop transcribes everything said in your videos and audio on its own, with Whisper on your computer, so you can read, copy and search it.
On this page
Polimake Desktop listens to your videos and audio and writes down what's said in them. It does this on your computer, with Whisper: the audio never leaves your machine and you don't need an account. Once a transcript is ready you can read each video, copy the text, jump to the exact moment of a sentence and find a video by what's said in it.
How it works
There's nothing to turn on. When Desktop indexes your folders, it transcribes every video and audio file as one more job, just like it makes thumbnails.
- It only listens to speech. A voice detector marks the stretches where someone is talking and Whisper transcribes only those. That way it doesn't invent sentences over noise or music.
- The language. It transcribes in the app's language, unless the speech is clearly in another one: then it detects it and uses it for the whole file.
- It works in parts. A long video is transcribed in ten-minute stretches, so it doesn't need much memory even if it runs for an hour.
- It's saved with the library. If you index again, nothing is transcribed twice.
The voice model
To transcribe, Desktop needs a Whisper model. It downloads by itself the first time there's something to transcribe; if you're offline at that moment, it tries again every few minutes.
You can also manage it yourself:
- Open Settings in the sidebar.
- Go to Models and pick the On this computer tab.
- In the Listening group you'll see the two available models:
- Whisper Base (about 148 MB): the one used when it's downloaded.
- Whisper Tiny (about 78 MB): lighter.
- Click Download on the one you want, or Delete to remove it. If you delete one, you can download it again.
Read a file's transcript
There are two places to see it.
In the details panel. Select a video or audio file in the explorer and open its details. When the transcript is ready, the Transcripts tab appears:
- Read the text sentence by sentence, each with its minute and second.
- Click a sentence to open the video right at that moment.
- Type in Search what's said to keep only the sentences with a word.
- Click Copy to take the whole text to the clipboard.
In the quick view. Open a video with the space bar and click What's said in the top bar. The transcript opens next to the video and follows what's playing; click a sentence to jump to it.
Find a video by what's said
Because every sentence is saved, search also finds what's said. Search everywhere or in a folder on your disk: videos and audio files where your words are spoken show up in the results with the moment on top of their thumbnail. When you open one, it starts at that moment.
What you may see in the meantime
| Message | What it means |
|---|---|
| Transcript pending | It's in the queue and will be done when its turn comes. |
| The voice model is missing | There's no model yet. Click View in Settings to download it. |
| No sound | The file has no audio track. |
| No speech | It has sound, but nobody is talking. |
| Couldn't transcribe it | Something failed. Click Try again. |
Frequently asked questions
Are my videos uploaded anywhere to transcribe them?
No. Transcription happens entirely on your computer with Whisper. Only the voice model is downloaded, once.
Does it cost anything?
No. It's part of Polimake Desktop, which is free, and you don't need an account to use it.
Which languages does it transcribe?
It uses the app's language by default and detects another language when the speech is clearly in it. It detects one language per file, not per sentence. When you generate subtitles in the video editor, you pick the language yourself: Spanish or English.
Which model should I use, Base or Tiny?
Base is the one Desktop prefers and the one it downloads by itself. Tiny takes half the space; download it if you're short on disk space.