🎙️ Transcription
Whisper Model Download Failed or Stuck: How to Fix It
Offline transcription needs the Whisper Large v3.1 model, a single file of about 3 GB (ggml-large-v3.1.bin). It's downloaded separately from the installer, either during setup or the first time you choose Offline / Local in the panel. Because it's large, a slow or unstable connection is the most common reason it fails.
Quick fixes
- Check your free disk space. You need at least 4 GB free on the drive where Captyne is installed (
C:by default). - Start the download again from the panel. Choose Offline / Local as the provider and accept the download prompt. Captyne downloads over several connections at once and continues from where it stopped.
- Turn your VPN or proxy on or off. Some networks block or slow large downloads. If you're behind a proxy, set it in Captyne's Settings → Network first. See Connection errors behind a VPN or proxy.
- Pause cloud sync and aggressive antivirus scanning while the file downloads, then try again.
Install the model manually
If the automatic download keeps failing, you can put the file in place yourself:
- Download the model in your browser from
https://assets.captyne.com/models/ggml-large-v3.1.bin. A download manager helps on slow connections. - Check the finished file is about 3 GB. A smaller file means the download was cut off.
- Move it to the models folder and keep the file name exactly
ggml-large-v3.1.bin:
- Windows:
C:\Captyne\models\ - Mac:
~/CaptyneRuntime/models/
- Restart After Effects and choose Offline / Local again. Captyne finds the model automatically.
On Mac you can also double-click Download Captyne Model on your Desktop to start the download.
Still not working?
If the model is in place but transcription fails, check that your graphics card is supported: see Offline transcription GPU requirements. Meanwhile, the Managed API provider works without the model.