Skip to content

Transcribe File(s)

What This Feature Does

Transcribe File(s)... provides queue-based transcription for existing audio/video files and writes per-file output session folders.

Design Rationale

Existing-file workflows prioritize throughput and repeatability:

  • queue multiple files/folders
  • monitor per-item status
  • generate consistent output artifacts
  • optionally apply LLM post-processing without changing source transcript

Queue and Status Model

Status Meaning
Queued Waiting to process
In progress Currently transcribing
Completed Finished with outputs available
Failed Processing error recorded

Input Handling

  • Supports drag-and-drop and Add Files... / Add Folder....
  • Folder input is expanded to supported media files.
  • Duplicate paths are de-duplicated in queue assembly.

Output Artifacts Per Item

Per input item, outputs can include:

  • transcript.txt
  • transcript.words.jsonl
  • transcript.srt
  • transcript.vtt
  • transcript.ttml
  • transcript.smpte-tt.ttml
  • transcript.llm.txt
  • transcript.llm-request.json

Choosing Output Formats

Use TXT for a readable transcript, SRT for broadly supported subtitles, and WebVTT for web players. TTML and SMPTE-TT are mainly for professional delivery requirements, while word timing JSONL is intended for custom tooling and detailed timing analysis.

The LLM output is a transformed companion to the original transcript, not a replacement for it. See Output Formats Explained for definitions and recommended bundles.

Settings Reference

Setting Notes
Output directory Root for per-item session folders
Chunking Controls transcript segmentation behavior, not audio recapture
Chunk (s) Segmentation interval for long-form readability
Subtitle cue split mode Affects subtitle pacing/structure
LLM post-processor Reusable provider, endpoint, model, prompt, timeout, credential, and wrapping configuration
LLM Post-Processor Output Sends each completed file transcript to the selected post-processor and writes transcript.llm.txt without changing the original transcript
Save LLM Request Payload Optionally saves the exact request body, which can contain transcript and rendered prompt content

Segmentation Principle

When chunking is enabled, the system segments transcript output for readability and subtitle utility. It does not re-record source media or change original audio fidelity.

LLM Post-Processing Path

If enabled:

  1. Base transcript is produced first.
  2. The selected library processor supplies its prompt, provider configuration, and optional wrapping tags.
  3. LLM processes the transcript.
  4. .llm.txt and optional .llm-request.json are written.

Choose the processor from the LLM Post-Processor menu. Edit Processor... opens that exact entry in the shared library, and Test LLM Post-Processing... stays live as the selection or processor settings change.

Editing a processor also affects any profile or Meeting workflow intentionally using the same entry. Duplicate it first when file batches need independent behavior.

Transformation Examples

Example 1: Base Transcript Output

Input media spoken content:

today we reviewed launch blockers and assigned action items

transcript.txt output:

today we reviewed launch blockers and assigned action items

Example 2: LLM-Processed Variant

Same base transcript with LLM enabled:

transcript.llm.txt output:

Today we reviewed launch blockers and assigned action items.

What changed:

  • punctuation and capitalization improved
  • meaning stayed the same
Use Case Chunking LLM Formats
Fast archive On Off TXT
Review/edit workflow On Off TXT + SRT/VTT
Analysis workflow On On TXT + SRT/VTT + LLM outputs

Common Pitfalls

Pitfall Symptom Fix
Huge queue + heavy LLM Long completion times Run base transcript pass first, LLM second
Output mismatch expectations Missing files Confirm toggles before run
Confusing chunking behavior Unexpected segment boundaries Adjust chunk seconds and cue split settings
Unexpected shared LLM behavior Another workflow changes after editing the processor Select or duplicate a separate processor for file transcription

Validation Checklist

  • Queue contains expected files only.
  • Status transitions complete without unexpected failures.
  • Output directory per item is created and populated.
  • LLM outputs exist only when enabled.

In-App Help Cues For This Feature

In Transcribe File(s)..., use ? help beside:

  • transcript/subtitle output toggles
  • reusable LLM processor selection and output behavior
  • cue split options
  • chunking semantics

Quick Tips

  • File transcription always preserves the source media. Chunking changes transcript segmentation, not the original audio or video.
  • The LLM result is a companion artifact. Keep transcript.txt enabled so you can compare a model rewrite with the recognized source.

See Also

Visual Notes

  • TODO: add screenshot of queue list and status chips.
  • TODO: add screenshot of file transcription settings panel.
  • TODO: add GIF adding a folder and processing multiple files.