UtterAI

Download started

Thanks — UtterAI is on its way.

Your download should have begun automatically.

Preparing your download…

Checking the latest release

Download again

Wrong platform? Windows installer · Debian / Ubuntu · Linux AppImage · Windows MSI

Install it

    First run

    1. Open UtterAI. A short intro explains what it does and lets you turn automatic update checks off if you'd rather they were.
    2. Pick a file. Choose an audio or video file, or drag one onto the window. MP3, WAV, M4A, FLAC, OGG, MP4, MKV, MOV and more.
    3. Choose how much. Transcribe the whole thing, or drag the handles to a range and do just those few minutes.
    4. Press Start. The built-in model needs no download, so the first transcription works offline straight away.
    5. Read, fix, export. Click a line to play it, double-click to correct a word, then export to TXT, SRT, VTT, Markdown or JSON.

    Verify what you downloaded

    Every release is built in the open by GitHub Actions, and the SHA-256 below comes from that build's own checksums.txt — not from anyone typing it in. If the hash of your file matches, the file is byte-for-byte what the workflow produced. This is optional; skip it if it isn't your sort of thing.

    —

    —

    Check it yourself:

    

        
    
          
            All checksums and signatures for this release →
          
        
    
      
    

      
    
        
    If something goes wrong
    
        
    
          Windows says "Windows protected your PC"
          
    
            UtterAI isn't code-signed yet — a certificate costs a few hundred
            pounds a year, and this is a free project. SmartScreen shows that
            warning for any unsigned installer. Click More info, then
            Run anyway. If you'd rather not, verify the SHA-256 above
            first: it proves the file is exactly what the public build produced.
          
    
        
    
        
    
          The AppImage won't start
          
    
            It needs the executable bit: chmod +x UtterAI_*.AppImage.
            On some minimal systems it also needs FUSE
            (sudo apt install libfuse2). Failing that, run it
            extracted: ./UtterAI_*.AppImage --appimage-extract-run.
          
    
        
    
        
    
          Audio won't play in the preview on Linux
          
    
            The window uses your system's GStreamer codecs for playback. The
            AppImage bundles them; a .deb install uses whatever your
            distribution has. If a file's preview is silent, install the codec
            packages:
            sudo apt install gstreamer1.0-plugins-good gstreamer1.0-plugins-bad gstreamer1.0-libav.
            Transcription itself is unaffected either way — it doesn't use
            GStreamer.
          
    
        
    
        
    
          Transcription is slower than I expected
          
    
            Speed depends on your CPU and the model. The built-in Base model runs
            several times faster than real time on a modern machine; Medium and
            Large (Turbo) are much more accurate and much slower. Transcribing a
            range rather than a whole recording is the quickest win.
          
    
        
    
        
    
          It says the model is unavailable
          
    
            The Base model ships inside the installer and is copied into place on
            first launch. If that copy failed — usually a full disk — open
            Settings, check the free space shown under Privacy &
            storage, and download a model from the list.
          
    
        
    
        
    
          Something else
          
    
            Open an issue with what you were doing and what happened:
            report a bug.
            Turning on Developer mode in Settings adds the technical
            detail to error messages, which helps a lot.