Enhancement: Vague "Transcription failed" error when you use your "Power mode" with deleted model

[benjaminmodayil]

Summary

The app showed a vague "Transcription Failed" message at the end of my recordings when I was using a power mode (auto switch based on context) due to a deleted local model

The history section showed a slightly more detailed error.

Transcription Failed: Failed to load the transcription model. Try selecting a different model or redownloading the current model.

---

Steps to reproduce

1. Set a model (Model A) in a power mode.
2. Delete Model A from local models.
3. Set a different model (Parakeet V3) as the default local model.
4. Record audio while using that edited power mode.
5. At the end of the recording you get a transcription error and the history shows the message above.

Expected behavior

• The error message should clearly indicate the particular Power Mode cause if possible (e.g., "Transcription failed because the power mode is configured to use a deleted model: Model A").
• Ideally, this could be also avoided for users if they had the ability to select a "Default model" in the model dropdown so whenever a model is deleted and a new default is set this won't break any workflows.

Proposed enhancements

1. Improve the error text when a model cannot be loaded because it was deleted or moved — make the message specific and actionable.
2. Add a "Default model" option when configuring power modes (in the model selection dropdown). This would allow:
   • Power modes have the option to use the current default model if the previously assigned model is removed.
   • Users to change their default model without breaking existing power-mode workflows.

Additional notes

• I only noticed this after a long recording (> 1 minute). We should error out on the recording initially if a model is selected but it doesn't exist. I can imagine a user recording an even lengthier transcription only to see an error after all that work.
• Making the error more specific or providing an automatic fallback would avoid transcription failures and improve the user experience.

[Beingpax]

Thank you for the feedback. Currently, even if the transcription fails, it remains in the transcript history so you can retranscribe it using different settings.

[ghzgod]

Thank you for the feedback. Currently, even if the transcription fails, it remains in the transcript history so you can retranscribe it using different settings.

This is not a solution. It's a workaround. I don't want to retry. It was working fine around eight to ten updates ago. I don't know what changed, but it's failing transcriptions every three or four. It should just either work, or retry on its own.