Dictionary

The Dictionary panel is where you teach Whiskers your shortcuts and your jargon. Two tabs, two purposes.

For the conceptual overview of when to use each, see Features → Vocabulary and snippets. This page covers the panel UI in detail.

Tab 1 — Text Replacements

Trigger phrases that expand into longer text after transcription. Whiskers replaces them before pasting.

Screenshot

Text Replacements tab showing a list of trigger-to-expansion pairs with toggle switches, hover-reveal edit/delete buttons, and an Add Replacement button

Adding a replacement

1. Click Add Replacement (dashed-border button at the bottom of the list).
2. A modal editor appears.
3. Fill in:
   • When transcribed as — the trigger you'll say (or that gets transcribed).
   • Replace with — the expansion that gets pasted instead.
4. Save. The new row appears at the top of the list.

Editing and deleting

Hover any row to reveal two icon buttons on the right:

• Pencil — opens the editor with the row's current values.
• Trash — deletes after a confirmation dialog.

Deletes are permanent. The confirmation shows the trigger so you can double-check before clicking through.

Toggle on / off

Each row has a toggle on the left. Off keeps the snippet saved but stops expanding it. Useful for temporarily disabling a snippet without losing the configuration.

CSV import / export

Click the ⋯ menu in the section header:

• Import CSV — bulk-add from a file with trigger,expansion columns. Use quotes for expansions containing commas. UTF-8 encoding.
• Export CSV — save your current list for backup or sharing.

Import shows a summary afterward: how many were added, skipped (invalid format), and ignored (duplicates).

Empty state

If you haven't added any replacements yet, you'll see an empty-state card with the icon and a one-liner explaining what replacements are for, instead of an empty list.

Tab 2 — Custom Vocabulary

A list of words Whiskers should treat as known. Helps with names, jargon, product terms, anything the model wouldn't otherwise recognize.

Screenshot

Custom Vocabulary tab showing a grid of word chips (Kubernetes, microservices, GraphQL, OAuth) with an add-words input at the top

Adding words

Type into the input at the top:

• Multiple words at once: separate with commas (Kubernetes, microservices, GraphQL).
• Press Add Words or hit return.

Each word appears as a chip below the input. A toast confirms the additions.

Removing

Click the X on any chip. A toast appears with an Undo button — accidental deletes are recoverable for a few seconds.

If a word is too long to display fully, hover the chip to see the full text in a popover.

What works well

• Add unusual spellings (your last name, niche product names).
• Add full phrases when ambiguous: "Q3 roadmap" beats just "Q3."
• Don't add common words — vocabulary helps most with terms the model doesn't already know.

See Features → Vocabulary and snippets for a fuller list of dos and don'ts plus example word sets for different fields.

CSV import / export

Same ⋯ menu as Text Replacements:

• Import CSV — single word column.
• Export CSV — current list for backup.

Import summary shows added / skipped / duplicates.

Where this matters

• Vocabulary affects transcription accuracy directly. For local models (Parakeet, Whisper), it's passed as a hint. For AI Enhancement, it's flagged so the LLM preserves the term rather than "correcting" it.
• Replacements run after transcription, before pasting. They never affect the raw transcript — only the final text that lands in your app.

You can use both together. Vocabulary keeps k8s recognized correctly; a snippet k8s → Kubernetes swaps it before paste.