Outlook Dictate Voice Commands: Delete, Select, Bold, and Beyond

The Outlook Dictate voice commands are a compact but real set, and most people use the feature as a one-way pipe: speak, accept what appears, then reach for the mouse to repair it. That habit leaves most of the tool's value on the floor, because the feature carries a compact but genuine command set for editing what your own voice just wrote. The commands fall into four families. Punctuation insertion, which dictates what your speech rhythm only implies. Deletion, which removes the last utterance or a selected stretch. Selection, which picks a phrase so you can act on it. And formatting, which applies bold, italics, or a new line. None of them replace a full sentence-level editor, but together they cover the majority of corrections a draft needs while your hands stay off the desk.

The mental model that makes the command set usable is the last-utterance marker. The engine tracks the most recent phrase you spoke, so commands like delete that remove exactly that phrase without you having to select anything. This is why the commands feel like conversation rather than surgery: you say a sentence, then say delete that, and the sentence vanishes. Long-form writers in particular benefit, because their drafts are long enough that reaching for the mouse mid-thought breaks the flow of composition. The command set is the difference between dictation as input and dictation as writing: one records your speech, the other lets you revise it while the thought is still warm.

There is a caveat that defines the boundary of the whole tool. The command set exists for the body of the message and follows the punctuation and formatting conventions of the engine's language. It does not extend to the recipient fields, the subject field, or the attachment zone, and it does not understand free-form instructions the way a chatbot might. Saying make the last line more formal is a wish, not a command; the engine will happily transcribe the wish as text. Learning the exact few phrases the engine accepts, and the exact boundary where they stop, is the fastest route to the streamlined drafting experience these tools promise.

The core phrases are few enough to memorize in one session, which is exactly the point. For deletion, the phrase is delete that or delete the last phrase, and it removes the most recent utterance, plus a few variants that operate on the selection. For selection, the command is select that, and the engine marks the last phrase so the next command has a target. Punctuation is the family that carries most of the daily value: say period, comma, question mark, exclamation mark, new line, and the symbols render themselves. Saying open quote and close quote is often smarter than hoping the engine infers them, and the same applies to parentheses. These are the phrases the tool was designed around, and they cover editing ninety percent of real drafts.

The rhythm that professionals adopt is punctuation-following, not punctuation-first. Speak the sentence, pause, then speak the punctuation command, so the marks land where the prosody intended. If you must apply punctuation while composing a long quote or a bulleted list, say the mark at the end of each clause and trust the engine to place it. The common mistakes are consistent: saying full stop in an English engine, which it may or may not accept, saying slash when you want a slash and getting the word written out, and forgetting that periods attached to abbreviations can become part of a sentence break. The pattern to build is speak, pause, command, continue.

The failure mode that surprises new users is the delete-without-target. If you say delete that after a moment of silence, or after the engine has paused its session, the target can be ambiguous, and the engine either deletes the wrong thing or asks for clarification it does not actually provide. The protection is a steady rhythm: keep the command adjacent to the utterance it modifies, and when in doubt, select that first to create a visible target. Read the transcript lines as they appear rather than after the fact, because repair by voice works best while the engine's memory of the utterance is short. Correction is a conversation, not a retrospective audit.

Formatting commands are where the command set grows from useful to expressive. After selecting a phrase with select that, you can say bold that or italicize that, and the engine applies the style to the marked text. The capitalization commands, such as capitalize that, are frequently forgotten and frequently invaluable, because the engine does not know a proper noun from a common one. New line and new paragraph are the structural pair that turn a flat wall of dictated prose into a readable message, and they are the formatting commands most people forget to use, so their emails arrive as one dense block even though the speech had natural breaks in it.

The ordering matters more than the vocabulary. Select first, then style: the engine must have a target before you give it a command. Saying bold that without having said select that leaves the engine guessing, and the result is either no change or styling the previous phrase. Get the rhythm right and the sequence runs in seconds: the sentence, the pause, select that, bold that, then the next sentence. For lists, dictate the items with new line between them, and if your version supports it, the bullet command frames each line. If the bullet command is missing, finish the lines and add the bullets in a one-pass keyboard movement after the dictation ends.

A caution that keeps formatting honest: the engine applies the style it knows, and the style it knows is button-equivalent, not pixel-perfect. A reference to the exact weight, the indent depth, or a colored highlight is beyond the phrase list, so heavy design work stays on the keyboard and mouse. What the voice handles is the semantics: this is bold, that is a fresh paragraph, here the list starts. The typographic finish is a human pass. In practice that division is ideal, because the semantic layer is where your flow lives and the visual layer is where your eye wants to be anyway. The dictation does the grammar of formatting, you keep the design.

Knowing precisely where the engine stops is what separates the fluent dictation user from the frustrated one. The engine's command list covers utterance-level operations; it does not cover whole-message operations. There is no voice command for save draft, send, add recipient, open the attachment picker, or change the reminder, and no amount of phrasing turns a wish into an action. The subject field, as always, is outside the list, and so is the table grid. The acceptance test is simple: if the operation needs the cursor to leave the body text, the keyboard or mouse owns it. The fastest users learn the boundary like a map and stop wasting speech on requests the engine will transcribe as letters.

The editing commands also have a version-dependence that deserves one honest note. Different builds of the client support slightly different phrase lists, and the help documentation plus a quick self-test at the start of a session is the durable way to keep your vocabulary current. A sentence is the test unit: say the sentence, then try delete that, and record what happens. The test takes two minutes and prevents mid-draft surprises. Similarly, the online speech service must be reachable for the commands to work, so a corporate proxy or a dodgy network can make the vocabulary appear smaller than it is. A failed delete command is sometimes a network issue wearing an editor's clothes.

The practical division of labor in a finished workflow looks like this. Voice handles the composition: the sentences, the punctuation, the deletes, the bold, the bullets. Keyboard handles the container: the subject, the recipients, the attachments, the send. That split is not a compromise; it is the natural separation of what speech is good at and what keys are good at. The writer who internalizes the boundary spends zero seconds attempting the impossible and all of the freed attention on the draft. The mouse becomes almost decorative, which is the quiet joy of a workflow where every tool knows its lane, and the editing phase stops feeling like a separate project.

The built-in phrase list is a starting point, and a tool such as Voiskey lets you extend it in two directions. First, macro-style phrases: create a custom word that fires a sequence of actions, such as a phrase that moves the cursor to the top of the body or a phrase that triggers a Quick Part. Second, dictation-level commands: custom terms that insert standard text with one utterance, such as a phrase that writes your standard status sentence. The design rule behind the extension is the same as the engine's: command phrases must be distinct from your ordinary vocabulary, or the tool will trigger them while you are writing a normal sentence. Pick words you rarely say, like push up or polish header.

The habit loop for growing the vocabulary is to add one command for each repeated action you catch yourself doing. When you notice yourself pressing Ctrl and Home for the third time, that is a macro candidate. When you notice yourself typing the same closing paragraph again, that is a snippet candidate. Keep a small scratchpad of the candidate phrases for a week, then build the top five. The quality bar is the sound of the trigger: distinct, short, and impossible to say accidentally. Over three months, the custom vocabulary turns your personal editing patterns into one-word operations, and the drafting flow condenses accordingly.

Discipline the library the same way you would a keyboard shortcut list. Document each phrase somewhere you will actually look, review quarterly, and remove the phrases you have stopped speaking, because a stale command library triggers confusion during a fast draft. Share the useful ones with teammates who dictate on shared mailboxes, so the vocabulary becomes a team asset rather than private lore. The boundary between the tool's official commands and your own additions is worth labeling in the documentation. The result is a vocabulary that fits your exact work, sits inside the tool's limits, and turns the daily draft into a spoken rhythm that your hands no longer interrupt.