Word Replacements
Create custom text snippets and automations.
How Word Replacements Work
Word Replacements allow you to define a word or phrase that will be automatically replaced with a different text snippet. This feature is primarily designed to correct speech recognition errors where the same word or phrase might be transcribed differently, but it can also be used for expanding abbreviations and inserting boilerplate text.
Word replacements are applied after the initial transcription and before AI enhancement, ensuring your text is corrected before any AI processing.
Use Cases
- Correct speech recognition mistakes: Speech-to-text systems sometimes transcribe the same word differently (e.g., "VoiceInk" might be transcribed as "Voicing", "Voice ink", or "Voiceing")
- Expand abbreviations: For example, you could have "brb" automatically replaced with "be right back"
- Insert boilerplate text: For frequent phrases, you can create shortcuts
- Standardize terminology: Ensure technical terms or proper names are always transcribed correctly
How to Use Word Replacements
- Go to the Dictionary tab in the sidebar.
- Select the Word Replacements section.
- Click the + button to add a new replacement.
- Enter the original word or phrase and the text you want to replace it with.
Adding Multiple Words at Once
You can specify multiple original words or phrases that should all be replaced with the same replacement text. This is particularly useful for correcting speech recognition variations:
- Separate with commas: Enter multiple originals separated by commas in the "Original Text" field
- All variants map to one replacement: Each comma-separated term will be replaced with the same replacement text
- Case-insensitive matching: Replacements work regardless of capitalization
Example: Voicing, Voice ink, Voiceing → VoiceInk
Practical Examples
Here are some real-world examples of how word replacements can correct common speech recognition issues:
Example 1: Correcting Product/Service Names
Original: github, git hub, get hub
Replacement: GitHub
Why: Speech recognition often mishears "GitHub" as "Git hub" or "Get hub"
Example 2: Technical Terminology
Original: api, a p i, aapi
Replacement: API
Why: Technical acronyms are frequently transcribed with spaces or misheard variations
Example 3: Proper Names
Original: chat gpt, chatgpt, chat g b t
Replacement: ChatGPT
Why: AI model names are commonly transcribed with spaces or pronunciation variations
Example 4: Expanding Abbreviations
Original: brb, be right back
Replacement: be right back
Why: Convert spoken abbreviations to their full forms for clarity
Important Notes
- Case-insensitive: Word replacements work regardless of capitalization. "GitHub", "github", and "GITHUB" will all be replaced if you specify "github" as the original.
- Word boundaries: Replacements respect word boundaries to avoid partial matches within other words.
- Processing order: Word replacements are applied after the initial transcription and before AI enhancement. This means that the replacement text will be included in the context provided to the AI, which can be useful for providing additional instructions or formatting.
Related
- Correct Spellings — teach the Whisper model to recognize specific words before transcription
- Filler Words — automatically remove words like "uh", "um", and "hmm"