File Transcription panel

Reference for the File Transcription window. For the conceptual overview (what it's for, when to use it, format support), see Features → File transcription.

Opening

Click the menu bar icon and choose File Transcription, or open Settings → File Transcription. Either way, the same window appears.

This feature requires a paid license. The window still opens on the free tier but is read-only — action buttons are disabled and a banner explains why.

Header

Title: File Transcription with the subtitle "Transcribe audio and video files with AI-powered processing and export options."

Top control bar — model and language

Sits above the drop zone. State varies:

Screenshot

File Transcription control bar with a Model dropdown set to 'Parakeet V2' and a Language dropdown set to 'English'

Model dropdown

Hierarchical menu:

• Local Models header → Parakeet V2, Parakeet V3, Whisper variants you've downloaded.
• A separator.
• Cloud Models header → providers (OpenAI, Groq, Deepgram, Google), each with their own submenu.

The currently active model is highlighted with a checkmark. Switching mid-queue applies to new jobs; jobs already in flight finish on their original model.

Language dropdown

• Auto-detect at the top.
• Alphabetical list of supported languages below.

Set this explicitly per file when possible — auto-detect adds processing time and occasionally misidentifies short utterances.

Drop zone

Big, friendly drag target with an icon and "Drag files here or click to browse."

Screenshot

File Transcription drop zone in highlighted state with files being dragged over from Finder

Behavior:

• Hover state — the border and background brighten while you drag over.
• Drop — files are validated and added to the queue.
• Click — opens the macOS file picker as an alternative to drag-and-drop.

File validation

Each dropped file gets:

• ✅ Valid — added to the queue, processing starts.
• ❌ Invalid — rejected with an inline error explaining why (unsupported format, unreadable file, etc.).

Invalid files don't enter the queue and won't block valid ones from processing.

Processing status

Just below the drop zone, a status line shows:

• Processing spinner + "X files processing" when one or more jobs are in flight.
• Empty / hidden when the queue is idle.

The file queue

Below the status line. One card per file, in the order added.

Screenshot

Queue of three file cards, one completed with green checkmark and expand arrow, one mid-transcription with a progress bar, one pending in gray

Each card shows:

• Filename and metadata — name, duration, size.
• Progress bar — animated during processing.
• Status indicator — pending dot, percentage during transcription, checkmark when done, X for errors.
• Expand arrow — opens the transcript view.
• Hover-reveal actions:
  • Retranscribe (circular arrow) — re-runs the job, useful after switching models.
  • Delete (trash) — removes the file with a confirmation.

Processing states

Expanded transcript card

Click the expand arrow on a completed file.

Screenshot

Expanded transcript card showing the full text, a waveform with a playhead halfway through, and an Export dropdown at the bottom right

Contents, top to bottom:

Transcript

Full text, selectable. Copy with Cmd+C like any other text.

Audio playback

• Play / pause button.
• Waveform — click anywhere to jump; drag the playhead to scrub.
• Duration timer.

The transcript scrolls in sync with playback when timestamps are available.

AI Transform section

Only appears if AI Enhancement is enabled.

• Preset buttons (Clean Transcript, Bullet Summary, Translate to Spanish, etc.) — click to apply.
• Custom Prompts link → opens the same prompt manager from Settings → AI Enhancement.
• A spinner appears while the transform runs.
• After the transform: transformed text replaces the displayed transcript, and a Revert button appears to restore the original.

Built-in prompts always run the cleanup baseline first (fix transcription errors, remove fillers), then apply the requested transform — so even a summary or translation starts from clean text.

Export menu

Dropdown with:

• Export as TXT — always available.
• Export as SRT — only when the model returns per-segment timestamps. Available for Whisper local, Groq (timed mode), Deepgram (timed mode), OpenAI whisper-1. Not available for Parakeet, Gemini, or OpenAI gpt-4o-transcribe variants.

Single-file export opens a macOS save dialog. For multiple files at once, select completed cards and use Export All in the queue toolbar.

Persistence

The queue survives:

• App restarts.
• Mac restarts.
• Switching to other settings panels.

Files stay in the queue until you delete them. There's no automatic cleanup.

Blocked state

When the free-tier limit is hit or your license isn't active, the window shows an inline banner explaining the reason, and the action buttons go disabled. You can still see your queue and read existing transcripts — only new processing and exports are blocked.

Common gotchas

• "Model setup required" — fix in Settings → Models first; File Transcription needs a working transcription model.
• Cloud API errors — verify the active key in Settings → API Keys for the cloud provider you picked.
• SRT button missing — your selected model doesn't return timestamps. Re-transcribe with Whisper local or Groq timed mode if you need subtitles.

More in Troubleshooting → Common issues.