Infinity Dictate Team
April 3, 2026 · 7 min read
The single biggest reason dictation feels slow and error-prone is not the recognition model — it is the user's workflow. Most people speak, see a mistake, stop dictating, reach for the keyboard, fix the error, then start speaking again. That interruption pattern breaks concentration, kills speed, and makes dictation feel worse than typing. It does not have to work that way.
Voice commands exist precisely to eliminate the keyboard from the correction and formatting loop. Once you internalize a small set of commands — punctuation, paragraph breaks, and correction phrases — you can handle everything by voice. The session becomes continuous. The output improves. And why dictation feels unnatural at first usually comes down to not knowing these commands exist.
Key Takeaways
- Saying punctuation by name ("comma", "period", "colon") inserts the correct character without touching the keyboard.
- Modern AI dictation infers sentence-ending periods and commas from natural pauses — explicit commands are most useful for less common marks.
- "Scratch that" and "delete that" are the two correction commands worth memorizing first — they cover most in-flow mistakes.
- Paragraph and line breaks are controlled entirely by voice with "new paragraph" and "new line".
- Building a command habit takes about one week of consistent use — after that, the commands become automatic.
Why Most Dictation Users Get Frustrating Results
The gap between a beginner dictation session and an experienced one is almost entirely about commands. A beginner speaks, sees a wrong word, and immediately stops to correct it manually. An experienced dictator speaks through the error, says "scratch that", speaks the correct version, and continues without losing momentum. The underlying recognition accuracy may be identical — but the experienced session feels faster and more fluid because corrections happen in-stream.
The same applies to punctuation. A beginner finishes a sentence, reaches for the keyboard to type a period, then resumes speaking. An experienced dictator either says "period" naturally at sentence end or — better still — trusts the AI to infer it from their speech cadence and never says it at all. That small difference, multiplied across hundreds of sentences, adds up to significant time savings and a qualitatively different experience. For more on the root causes of dictation friction, see our guide on why dictation feels unnatural at first.
Punctuation Commands: The Foundation
Punctuation commands are the entry point for hands-free dictation. The pattern is simple: say the name of the punctuation mark, and it appears in the text. These commands work in virtually every AI dictation system:
Basic punctuation: "period" or "full stop" inserts a period. "Comma" inserts a comma. "Question mark" inserts a question mark. "Exclamation point" or "exclamation mark" inserts an exclamation point. "Colon" inserts a colon. "Semicolon" inserts a semicolon.
Paired marks: "Open parenthesis" and "close parenthesis" insert parentheses around a phrase. "Open quote" and "close quote" insert quotation marks. "Dash" inserts an em dash. "Hyphen" inserts a hyphen between compound words.
Apostrophes: In most dictation systems you do not need to say "apostrophe" for contractions — the system recognizes "don't", "can't", and "it's" from spoken context and inserts the apostrophe automatically. For possessives, the same applies: saying "the company's revenue" produces the correct output without any explicit command.
The important practical note: for common marks — periods at sentence ends, commas after introductory clauses — you often do not need to say anything. AI dictation infers these from natural speech patterns. Explicit punctuation commands are most valuable for colons, semicolons, dashes, and parentheses, which are harder to infer from context. Understanding how AI dictation accuracy works explains why inference is reliable for common patterns but less so for rare ones.
Formatting Commands: Paragraphs, Capitalization, and Lists
Formatting commands control the structure of the text — where paragraphs begin, when capitalization changes, and how lists are created. These commands transform raw dictated prose into structured text without a single keystroke.
Paragraph and line breaks: "New paragraph" inserts a paragraph break (equivalent to pressing Enter twice in most contexts). "New line" inserts a single line break (equivalent to Shift+Enter). In practice, natural long pauses in speech also trigger paragraph breaks in most AI dictation systems, so you may find you rarely need to say "new paragraph" explicitly.
Capitalization: "Caps on" activates all-caps mode for a stretch of text. "Caps off" returns to normal. "Cap" before a word capitalizes just that word. Most systems also automatically capitalize the first word after a sentence-ending period, and proper nouns it recognizes from context.
Lists: "New bullet" or "bullet point" starts a new bulleted list item. For numbered lists, saying "number one" followed by the item content, then "number two", and so on produces a sequential list. Some AI dictation tools also recognize sequential spoken structure — "first", "second", "third" — and auto-format the result into a list during an AI polish pass.
Correction Commands: Fix Mistakes Without the Keyboard
Correction commands are the most important commands in any dictation workflow. They are what make hands-free editing possible and what allow you to maintain dictation flow even when the recognition makes an error.
"Scratch that" — deletes the last phrase or utterance spoken. This is the single most valuable command to memorize. If the previous sentence came out wrong, say "scratch that" and the entire phrase disappears. Then speak it again correctly. This is faster than any keyboard correction because it does not require moving your hands or your eyes to the cursor.
"Delete that" — removes the last word. Use this for single-word errors when the rest of the sentence was correct.
"Correct [word]" — targets a specific word for replacement. Speak the word you want to correct, then speak the intended word. Not all dictation systems support this command in exactly the same form, but most support some variant of targeted word correction.
"Undo that" — reverses the last action, including the last correction. Useful if "scratch that" deleted more than intended.
The key behavioral shift with correction commands is to keep speaking. When you hear a wrong word, do not stop — finish the sentence, then correct. Stopping mid-sentence interrupts the AI's language model context and often produces worse recognition for the resumed phrase. Finish the thought, correct in the gap, move forward. For a deeper treatment of correction workflows, see our guide on fixing dictation errors efficiently.
Navigation and Editing by Voice
Beyond corrections, some dictation systems support navigation commands that let you move through the document by voice. Support for these varies more than punctuation or correction commands, but they are worth knowing.
Cursor movement: "Go to beginning of line", "go to end of line", "go to beginning of document", and "go to end of document" move the cursor to those positions in many dictation tools. "Move up [number] lines" and "move down [number] lines" navigate by line count.
Selection: "Select [word]" highlights a specific word. "Select all" selects all text. "Select last [number] words" selects a range of preceding words for replacement or deletion.
Deletion: "Delete previous word" removes the word before the cursor. "Delete next word" removes the word after. "Delete line" removes the entire current line.
A practical approach: learn the correction commands first (they work everywhere), then add navigation commands as your system supports them. Trying to learn all commands simultaneously creates unnecessary cognitive load and slows the habit formation that makes commands automatic.
Building a Command Habit
Command fluency follows a predictable curve. In the first two or three sessions, you will forget the commands exist and reach for the keyboard out of habit. That is normal. The transition happens when you catch yourself reaching for the keyboard and say the command instead. After about a week of consistent dictation, the core commands — "scratch that", "new paragraph", "period" — become automatic enough that you stop thinking about them.
A practical approach: start with just three commands. "Scratch that" for the last phrase, "new paragraph" for structure, and one punctuation command for whatever mark you use most. Dictate at least 200 words per session using only those three commands. After three sessions, add two more. By the end of the first week you will have internalized the core set without overwhelming the learning curve.
The compound benefit of command fluency is that your dictation speed increases not just from faster corrections, but from reduced interruptions. Each keyboard reach-and-correction costs 3–5 seconds of momentum and concentration. Eliminate those interruptions and the session feels qualitatively different — more like thinking out loud than operating software.
When to Type Instead of Speaking Commands
Voice commands are not always the right tool. There are specific situations where typing is genuinely faster, and knowing when to switch prevents friction from going in the wrong direction.
Complex structural edits: If you need to rearrange multiple paragraphs, insert a table, or restructure a document significantly, the keyboard and mouse are faster. Voice commands are optimized for linear, forward-moving composition — not for document-level reorganization.
Proper noun heavy passages: If you are dictating a passage dense with unusual names, technical terms, or specialized jargon that your system regularly misrecognizes, it may be faster to type those sections and dictate the surrounding prose. The correction overhead for repeated misrecognitions exceeds the dictation speed benefit.
Quiet or sensitive environments: Some situations genuinely require silent work. In those cases, the keyboard is the right tool and forcing voice commands creates social friction that outweighs the speed benefit.
Short one-off edits: If you need to correct a single character in a long document — fix a capital letter, change one digit — the keyboard is faster. Voice commands shine in continuous composition, not in point corrections on already-existing text.
The goal is not to eliminate the keyboard entirely but to eliminate it from the composition loop — the active drafting phase where voice commands deliver their greatest advantage.
Frequently Asked Questions
How do you add a period when dictating?
Say "period" or "full stop" at the end of a sentence. Most AI dictation systems also infer sentence boundaries automatically from natural pauses, so you may not need to say "period" at all once you speak in complete sentences with natural pacing. For question marks say "question mark", for exclamation points say "exclamation point" or "exclamation mark".
How do you start a new paragraph when dictating?
Say "new paragraph" or "new line" to move to the next line. "New paragraph" typically inserts a double line break (paragraph spacing), while "new line" inserts a single line break. In most dictation software, a long natural pause also triggers a new paragraph automatically.
How do you correct a mistake without stopping dictation?
Say "scratch that" to delete the last phrase spoken. Say "delete that" to remove the last word. For specific corrections, say "correct [word]" and speak the replacement. These commands let you fix errors in flow without reaching for the keyboard, which preserves your dictation momentum and keeps the session continuous.
Can you dictate bullet points and numbered lists?
Yes. Say "new bullet" or "bullet point" to start a list item in most dictation tools. For numbered lists, say "number one", pause, speak the item, then "number two", and so on. In AI-polished dictation, speaking items in sequence with "first... second... third..." will often auto-format into a list after the polish pass.
Do you have to say every punctuation mark when dictating?
Not necessarily. Modern AI dictation infers common punctuation from speech patterns and pauses. Commas after introductory phrases, periods at sentence ends, and question marks after rising intonation are often inserted automatically. Explicitly speaking punctuation is most useful for colons, semicolons, dashes, and parentheses — which are less predictable from context alone.