MachinesFluent Docs
Troubleshooting

Model Download Troubleshooting

Fix local speech model downloads, stuck model loading, and cache issues.

Local speech engines may need to download model files before first use. Keep MachinesFluent open during the first download and avoid switching networks while a model is being fetched.

What A Model Download Means

MachinesFluent uses local model files for local speech recognition. The installer may not include every model file, so the app can download a selected model when you choose it or ask it to download.

Engine familyWhat to expect
ParakeetCPU-friendly local model download.
MedASRCPU-friendly local model download with medical vocabulary focus.
Faster WhisperLarger local model files; best experience may depend on hardware and selected model size.
Kyutai MoshiLarge local model files; best experience may require compatible NVIDIA GPU hardware.

Cloud speech engines do not use these local model downloads, but they require network access and a provider setup.

If Download Progress Looks Stuck

Try this first:

  1. Leave MachinesFluent open for several more minutes, especially near the end of a large download.
  2. Check that the network is still connected.
  3. Avoid repeatedly clicking the same download control.
  4. Restart MachinesFluent only if there has been no visible change for a long time.
  5. After restart, reopen speech settings and check whether the model is already available.

Some downloads verify files after the visible transfer. A progress indicator can appear stuck near completion while final file checks finish.

If A Download Fails

Common causes include:

  • network interruption
  • not enough disk space
  • security software blocking the model cache
  • Windows symlink permission warnings
  • a partially downloaded cache from an interrupted first attempt

Recommended recovery:

  1. Confirm you have enough free disk space.
  2. Restart MachinesFluent.
  3. Try the download again on a stable network.
  4. Try a smaller or CPU-friendly local model first.
  5. If the same model repeatedly fails, contact support with logs.

Some model-download tooling uses symlinks in the local model cache. On Windows, that can trigger permission warnings if Developer Mode is not enabled.

If MachinesFluent says the model is available after the warning, you can usually continue. If the model still is not available, enable Windows Developer Mode, restart MachinesFluent, and retry the download.

Developer Mode is only for Windows file handling

Enabling Developer Mode is a Windows permission workaround for model-cache symlinks. It is not a MachinesFluent account setting and it does not change your speech provider choice.

If The Model Loads Slowly Every Time

Repeated slow loading can mean the model is still being downloaded, the selected model is too heavy for the machine, or the local cache was not detected cleanly.

Try:

  • using Parakeet or MedASR as a baseline
  • closing GPU-heavy apps before loading GPU models
  • switching away from the current model and back after the download completes
  • restarting MachinesFluent after a completed first download

When To Contact Support

Include:

  • MachinesFluent version
  • selected speech engine and model
  • whether the issue is a first download or repeated loading
  • approximate free disk space
  • whether Windows showed a symlink or Developer Mode warning
  • relevant log files from %APPDATA%\MachinesFluent\logs\

On this page