FAQ and Troubleshooting

Use this page when the main workflow is clear but one part is still blocking you.

Which desktop build should I install?

• Most users should install the latest release build from Desktop downloads or GitHub Releases, not a source build.
• Pick the build that matches your operating system, CPU architecture, and preferred installer format.
• Source builds and development commands belong in the project README.

Sona keeps asking me to finish setup

• Open the onboarding banner and complete the model and microphone steps.
• If you skipped setup earlier, make sure both Live Record Model and Batch Import Model are configured in Settings > Model Settings.
• If you hid the reminder banner, reopen setup manually from the banner flow or settings-related entry points when Sona prompts again.

Download Recommended Models fails or never finishes

• Keep Sona open and stay on a stable network; the recommended model package can take time to download and extract.
• Check available disk space, then retry from Settings > Model Settings if onboarding was interrupted.
• If it repeatedly stalls, open Diagnostics and review the model download or local runtime message before retrying.

A model is downloaded, but Sona still says it is missing

• Open Settings > Model Settings and confirm the right slot is selected: Live Record Model for live transcription or Batch Import Model for file transcription.
• Make sure the model folder still contains the extracted files. Moving or renaming model folders after download can make Sona lose the path.
• If the model was downloaded outside Sona, choose it manually in Settings > Model Settings instead of relying on first-run setup.

Live Record does not start

• Check microphone permission in your operating system.
• Confirm that a Live Record Model is configured.
• Make sure the input source is correct: Microphone or Desktop Audio.

Live Record starts but still cannot hear the microphone

• Open Settings > Input Device and select the microphone you actually want Sona to use.
• Check operating-system microphone permission and make sure another app is not holding the device exclusively.
• If the current source is Desktop Audio, switch back to Microphone to confirm the input path.
• Open Diagnostics if the selected device looks correct but Sona still reports no input.

Where can I check local transcription diagnostics?

• Open Settings > General, then select Diagnostics.
• Use it to inspect local transcription readiness, model paths, input device state, and the packaged runtime environment.
• When asking for help, share the visible diagnostic message or error text instead of guessing from the symptom alone.

Batch Import does not start

• Make sure a Batch Import Model is configured.
• Confirm that the file extension is supported.
• If Sona reports an unsupported format, convert the file first and try again.

LLM Polish or Translate is disabled or fails

• Confirm that Settings > LLM Service has the correct provider credentials.
• Make sure the feature itself has a model assigned: Polish Model for polishing, Translation Model for translation.
• Use Test Connection before retrying.
• If you use a custom endpoint or local service such as Ollama, verify that service first.

I cannot find Auto-Polish

• Open LLM Polish > Advanced Settings for Auto-Polish, frequency, keywords, directive polish modes (Clean, Verbatim, Formal), and custom context.
• If you want the map of where those controls fit, continue to Vocabulary and Advanced Settings.

Export only shows Original

• Translation and Bilingual only appear when the transcript already contains translation text.

Speaker Review is empty or has no candidates

• Speaker Review groups existing speaker metadata. If the transcript was created without speaker attribution, there may be nothing to review.
• Candidate suggestions depend on Speaker Profiles in Settings > Vocabulary, imported reference samples, configured speaker models, and the profiles enabled for the current project.
• If a speaker badge is visible in the editor, you can still click it and assign a profile manually.

I cannot find Version Snapshots

• Version Snapshots only appears for saved project items that already contain transcript segments.
• It is hidden for the temporary current transcript and for live recording drafts that are still in progress.
• Snapshots are created before bulk rewrite operations such as LLM Polish, Translate, Re-transcribe, and before restoring from another snapshot.

A restored backup opens text but no audio playback

• Backup archives are intentionally lightweight. They include config, project data, light history transcripts and summaries, automation state, and dashboard LLM usage.
• Original audio files are not included, so restored entries may reopen for reading and editing but lack playback until the source audio is available through another path.

Why do new items appear in Inbox first

• Inbox is the default holding area for recordings and imports that are not assigned to a project yet.
• Open a specific project before starting Live Record or Batch Import if you want new items to stay in that project automatically.
• You can also move saved items later from Inbox into a project from Projects and Inbox.

What happens when I delete a project

• Deleting a project removes the project itself, but it does not delete the recordings or imports that were inside it.
• Sona moves those items back to Inbox, so you can reassign or continue using them later.

Why do I see a Draft item while I am still recording

• During an active live recording, Sona can create a visible Draft item so the session already has a saved place in Projects.
• When you stop recording, Sona completes that same item instead of creating a second saved copy.
• Treat the draft as the live session while capture is still running; the finished transcript stays on the same entry.

Live Caption does not appear

• Go back to Live Record and make sure Live Caption is turned on there. Settings > Subtitle Settings only controls the window behavior and appearance.
• If you only want floating system-audio subtitles, you do not need to start recording first. Turning on Live Caption is enough.
• Live Caption depends on the same offline live transcription setup, so make sure a Live Record Model is configured.
• If you want the split between the page toggle and the settings layer explained first, continue to Live Caption and Voice Typing.

Voice Typing does not work

• Turn on Voice Typing in Settings > Voice Typing.
• Confirm that the page shows the shortcut, model, VAD, input device, and readiness state you expect.
• Make sure a live transcription model is configured, because Voice Typing depends on the same offline transcription setup.
• If you still need the setup path or the mode choice explained, continue to Live Caption and Voice Typing.

Where do I recover interrupted batch or automation work

• Open the header notification center if Sona reports pending recovery items.
• Use Recovery Center to resume or discard interrupted batch and automation work.
• If you were looking for diagnostics or backup instead, go to Settings > General.

Playback controls are missing

• The audio player only appears when the current transcript has an audio source available, such as a saved recording or processed file.

What if I forget my Master Password for Cloud Sync (E2EE)?

• If you saved the Emergency Recovery Key generated during vault setup, you can use it to reset access.
• If both the Master Password and Emergency Recovery Key are lost, no one (including developers) can decrypt or recover the encrypted data on the sync server due to zero-knowledge end-to-end encryption. However, transcripts and audio already stored locally on your device remain unaffected and fully accessible.

Does Cloud Sync upload my audio recordings to the cloud?

• No. Cloud Sync only synchronizes transcript segments, summaries, project metadata, and preferences. All original audio recording files remain stored locally on your machine and are never uploaded to WebDAV or S3-compatible cloud storage.

Why does Cloud Sync show a conflict?

• When multiple devices edit the same transcript offline or make concurrent edits within sync intervals, Sona detects the discrepancy and flags it in Settings > Cloud Sync > Conflict Center.
• In Conflict Center, you can review diffs side by side and manually choose between keeping the local version, remote version, or merging them.

How do I recover an accidentally deleted transcript?

• Deleted transcripts are not wiped from disk immediately; they move to Trash in the Projects rail.
• Switch to the Trash scope, select the item, and click Restore to return it to its original project or Inbox.
• Data is only permanently removed from disk when you click Empty Trash or select permanent deletion.

Can I cancel an in-progress batch transcription job?

• Yes. Click the Message Center icon in the header to view active batch transcription queues, where you can cancel active tasks in real time.

What if media import fails or reports a missing decoder?

• Sona uses built-in media decoding by default. If you encounter non-standard audio/video containers or missing decoders, go to Settings > Storage Management and specify your installed FFmpeg executable in Custom FFmpeg Path.

I want to build or develop Sona

• Use the project README for source builds and development commands.
• If you only want to use Sona, install a release build instead; source builds are mainly for contributors and development work.