Wispr Flow Dictionary: Fix Names and Use Spoken Snippets

Imagine dictating a note about a fictional colleague, Nera Vellorin, and having to fix the name again. Later, you need to insert a review reply you've already written. Both problems involve repeated wording. They need different settings.

The Wispr Flow dictionary supplies vocabulary for recognition. If the same wrong spelling keeps appearing, an explicit correction pairs it with the spelling you want. For an entire saved reply, use a snippet triggered by a spoken phrase. Decide which problem you have before adding entries.

This guide covers documented Mac and Windows setup. The names and saved reply are hypothetical inputs; the checks are yours to run. Optijara has not installed Flow or performed a microphone session for this article, so these examples provide no evidence of recognition accuracy.

Choose the right fix: Word, Correction, or Passage

A recognition hint is not a replacement rule

Word, Correction, Passage is Optijara's editorial framework for making this choice. It is not an official Wispr feature or a validated method. The principle is simple: make the smallest change that addresses the problem you've observed.

The dictionary documentation separates vocabulary from persistent-misspelling corrections. Neither feature promises verbatim transcription of a whole passage. That distinction matters before you spend time tuning entries.

Use a snippet when you need saved wording

According to the snippet guide, a spoken trigger inserts saved text. Treat that text as a draft. A review request still needs someone to perform the review, and inserting it does not send a message or execute a command.

The diagram maps this editorial decision. It does not describe Flow's internal processing.

Add a recurring name to your personal dictionary

Start with harmless fictional vocabulary

Assume Flow is installed, you are signed in, and dictation already works in a disposable document. Use a personal entry for this exercise. Nera Vellorin and Tessera Lab are fictional names, included to make the instructions concrete without exposing confidential information.

Add the colleague's name first. If you change several entries at once, it becomes harder to tell which change mattered when you inspect the next dictation.

Add the word, then inspect fresh dictation

Use the documented desktop sequence:

1. Open Flow and select Dictionary in the sidebar.
2. Select Personal if needed.
3. Choose Add new.
4. Enter Nera Vellorin as the desired word or phrase.
5. Select Add word.

Now try dictating, 'Nera Vellorin will review the Tessera Lab draft.' Read the name in the destination document. The sentence is a proposed input, not a transcript captured for this guide.

Don't assume an edit will teach Flow a word automatically. The same dictionary guide says desktop auto-learn words must be enabled and dictated text must be edited. Even then, not every correction becomes an entry. Exclusions include file paths, URLs, code symbols, keys, hashes, and IDs. For a recurring technical identifier, inspect the result rather than counting on automatic learning.

Turn a persistent misspelling into an explicit correction

Enter the desired spelling before the wrong spelling

Suppose you repeatedly observe Nera Velorin where you intended Nera Vellorin. This is a hypothetical error, not an observed Flow result. It illustrates when an explicit correction would be worth trying.

In Dictionary, choose Add new and enter Nera Vellorin. Turn Correct a misspelling on, enter Nera Velorin in the wrong-spelling field, then select Add word. Check the direction before saving: wrong Nera Velorin should become desired Nera Vellorin.

The correction instructions put this option in the new-entry operation. Each entry holds one correction, which can later be changed. Editing an ordinary word is not a way to create several replacement rules.

Already added the personal name above? The same guide says desktop Add word upgrades an existing personal word without a correction when Share with team is off. This includes imported or auto-learned words, but not team words. You do not need to delete the personal word first. This upgrade does not apply on iOS.

Keep replacement rules narrow and inspect conflicts

If an existing entry prevents saving, look through the relevant personal and shared lists. The dictionary guide says dictionary search covers the selected tab. Check the intended library and consult the documented conflict guidance before deleting anything.

The correction rules match whole words without regard to capitalization and prefer longer overlapping phrases. Replacements are literal, not language-aware. Avoid mapping wording you sometimes want unchanged; surrounding text can affect spacing and punctuation.

Repeat the proposed name sentence. Then dictate an ordinary sentence that should remain unaffected. Record what actually appears; the presence of a saved correction does not establish that it worked.

There is another boundary here. The dictionary guide says free-form and custom prompt requests run without your dictionary. More dictionary tuning will not give that feature control over custom-prompt rewriting.

Create a spoken snippet for a reusable review reply

Choose a distinctive trigger

For the fictional task, use insert review request. It is more deliberate than a common word such as 'thanks.' Still, consider whether you might say the phrase naturally when you don't want an expansion.

Following the official snippet guide, open Snippets, a separate sidebar page from Dictionary. Select Personal if a picker appears, then choose Add new. Enter the trigger in Snippet and put this saved wording in Expansion:

> Thanks for the draft. Please identify claims that need evidence and leave your questions for the next revision before we approve it.

Select Add snippet. This is proposed wording for you to save, not a fabricated expansion result. The guide says personal snippets are available on every plan, including Basic.

Save static text and review the insertion

Speak the complete trigger into a disposable document. Try it alone, then within a sentence. Read the entire insertion and check how it fits the surrounding text.

The snippet documentation says expansions keep their saved casing, though Flow can adjust surrounding capitalization or punctuation. On desktop, snippets contain static text; dynamic variables such as today's date are unsupported. This saved reply won't fetch a draft or complete an AI review. You decide whether its wording belongs in the conversation.

For desktop triggers inside sentences, internal punctuation must match, and the trigger cannot touch letters, numbers, or underscores. The matching guidance says the longest conflicting trigger wins. Keep your example distinct and inspect both standalone and in-sentence insertion.

If personal and team snippets share a trigger, the personal entry takes precedence regardless of capitalization. Check both libraries before concluding that the team wording has disappeared.

Run the Word, Correction, Passage check before relying on it

A small implementation checklist

A saved setting is weak evidence. The text in your destination document is what matters. Work through this checklist after each change, keeping enough detail to decide whether the entry deserves to stay.

For the fictional example, inspect the name first. Add the correction only if your observations justify it. Check the whole saved reply separately, including whether it makes sense where you inserted it.

This JSON records the editorial framework. It is not an import file or executable configuration, and it contains no test results.

{"framework":"word-correction-passage","word":"recognition-vocabulary","correction":"wrong-spelling-to-desired-spelling","passage":"static-snippet","verification":"inspect-inserted-text","execution":"examples-unexecuted"}

Record observations without inventing an accuracy score

Keep a blank observation note and fill it only after checking the inserted text:

Platform: ____; entry type: ____; intended wording: ____; actually observed wording: ____; destination app: ____; next action: ____.

The measurement plan below tells you what to record. No observations or benchmark results have been collected for this article.

A correctly expanded review request can still be inappropriate for a draft that's already approved. Checking the words and deciding whether to send them are separate jobs.

Common mistakes: too much learning, too broad a trigger

Do not confuse recognition, rewriting, and automation

The easiest mistake is expecting every edit to become a learned word. Automatic learning has conditions. Another is expecting dictionary entries to govern custom prompts, despite the documented exclusion. And a snippet that sounds like an instruction still inserts text; it does not carry out the instruction.

The Raycast clipboard and snippets guide covers a different input mechanism for prepared text, including clipboard or AI handoffs. It is useful follow-on reading, but its settings don't apply to Flow recognition or spoken triggers.

Keep the personal example out of shared libraries

Use harmless sample wording. Don't put passwords, API keys, real client addresses, or confidential names into this exercise. Calling a library personal does not establish a privacy assurance.

The desktop management guide says sharing adds the team entry before removing the personal copy. A failed removal can leave both copies; inspect the result rather than treating sharing as a backup. Keep this demonstration personal.

The same guide warns that deletion is permanent and neither desktop page offers export. Preserve harmless sample wording separately before removal. The dictionary documentation also says automatic learning will not re-add a deleted word; add it manually if needed. Inspect the entry and library before confirming deletion.

Platform limits and a safer troubleshooting path

Mac and Windows first; mobile is not identical

The dictionary guide says Android can use corrections synced from desktop or iOS but cannot create them. Follow supported-platform instructions for creation; these desktop steps are not an Android walkthrough.

For historical context, 9to5Mac's iPhone account describes a keyboard-specific workflow. It does not establish current feature parity or pricing. Use current official instructions for setup.

Check account and sync before escalating

According to the dictionary guide, words and corrections sync online across devices using the same account. For a missing word, check the account and selected library, then the desktop refresh control.

For a snippet that fails to expand, check the complete trigger. The snippet troubleshooting guide suggests restarting desktop Flow to retry synchronization. That is a troubleshooting step, not a guarantee of recovery.

Avoid signing out as your first response. On iOS and Android, it clears the local dictionary until successful synchronization, according to the dictionary guide. The snippet guide says signing out or switching accounts also clears Android's local snippets. None of these sync behaviors establishes offline transcription support.

An organization policy can hide a page and block changes while saved words still apply and snippets continue expanding. A missing page does not prove the stored behavior was erased.

If the problem persists, prepare a non-sensitive support report. Include the platform and expected versus actual wording, plus the relevant word or trigger where safe. Add any error message and the steps already tried. Leave mobile editing outside this desktop exercise.

Keep a small vocabulary and a deliberate review habit

A short library you can explain is preferable to a long one you no longer trust. Start with the fictional name and harmless reply. Add a correction only after observing a recurring error, and expand the collection when each entry has a clear purpose.

Read a reply in context before it becomes a commitment. The Granola guide to confirming actions before sharing addresses that judgment in meeting notes, a different workflow from desktop dictation.

If recurring vocabulary and review wording are scattered across your team's writing tools, Optijara can help map the workflow and design a small evaluation.