Auto-Editing
What is auto-editing?
Auto-editing is an optional post-processing step that runs after Whisper transcribes your dictated audio. Before your text is inserted, GPT-4 cleans it up by:
- Removing filler words — um, uh, like, you know, sort of, kind of, basically, right, so, well
- Fixing grammar errors introduced by natural speech patterns
- Adding punctuation — commas, periods, and capitalisation where appropriate
- Smoothing out false starts and repeated words
The result is polished, readable text suitable for documents, emails, or messages — without you having to say "period" or "comma" while speaking.
Enabling or disabling auto-editing
Auto-editing is enabled by default. To toggle it:
- Open the Settings tab.
- Find the Auto-Edit toggle.
- Toggle it on or off.
- Click Save Settings.
Changes take effect for the next dictation session.
When to disable auto-editing
Auto-editing is designed for natural dictation. You may want to disable it if:
- You need the raw transcript without any changes (e.g., for verbatim records or debugging).
- You are dictating code or structured content where filler word removal could corrupt the output.
- GPT-4 is making unwanted edits to technical terminology or proper nouns.
- You want to minimise API calls and costs (auto-editing uses an additional GPT-4 API call).
How auto-editing works
- Whisper returns the raw transcript text.
- The text is sent to GPT-4 with a prompt instructing it to clean up the transcription.
- GPT-4 returns the edited text.
- The edited text is inserted at the cursor.
This process adds a short delay (typically 1–3 seconds) compared to inserting the raw Whisper output directly.
API usage
Auto-editing uses a separate GPT-4 API call for each dictation session. This contributes to your OpenAI API usage. Check your OpenAI account dashboard to monitor costs if you dictate frequently.