Formatting pass
Also: post-processing
Updated
Definition
The formatting pass is the step after speech recognition that turns a raw transcript into text a person would have typed: punctuation, capitalization, numbers as digits, fillers removed, and a tone suited to where the text is going. It is where most of the difference between dictation tools lives.
A recognizer produces words. It does not know that "ten thirty am" should be 10:30 AM, that a Slack message is lowercase and a client email is not, or that "bob at acme dot com" is an address. The formatting pass knows, or is supposed to. Early tools did it with rules; the current generation runs a small language model over the transcript; the best do both, with rules for anything that has one right answer and the model for judgment.
It is also the step with the most ways to go wrong. A model asked to "clean this up" will reword, tighten, and occasionally invent, and a slow model turns a fast tool into a waiting room. How a product constrains its formatting pass tells you more about its quality than which recognizer it uses.
What a good pass does
Removes fillers and false starts. Punctuates and capitalizes. Writes numbers, times, dates, prices, units, emails, and URLs in their proper form. Applies spoken commands like "new paragraph". Inserts snippets and dictionary terms exactly as saved. Matches the tone of the destination app, casual or professional, and can be pinned per app. Leaves terminals and code editors untouched. Everything on that list is either a rule or a narrowly briefed model, and the order matters: rules first, so the model never gets to second-guess a clock time.
Rules versus the model
A language model is good at judgment and unreliable at precision. Given "ten thirty AM" it usually writes 10:30 AM, and "usually" is the problem, because you will not notice the time it wrote 10:30 PM until the meeting is missed. Anything with one correct answer, clock times, spoken punctuation, dictionary words, snippets, symbol names, should be handled by a deterministic rule that produces the same output every time and can be tested. The model gets what is left: sentence flow, paragraph breaks, which fillers were filler, and tone.
Guardrails
Three are worth insisting on. A content check: if the model’s output loses a word you said or adds one you did not, discard it and keep the rule-formatted transcript. A time budget: if the model has not answered within a fixed window, the transcript lands without it. And a verbatim escape: apps where any rewrite is a bug get no model at all. With those in place, a formatting pass can only improve the text or leave it alone; it can never break a dictation.
How to format text yourself with Apple Dictation
Apple’s built-in Dictation does a small part of this work automatically and leaves the rest to spoken commands. Knowing where that line sits tells you what a fuller formatting pass would be doing for you.
The automatic part is auto-punctuation, which in supported languages inserts commas, periods, and question marks as you speak. Paragraphs, capitals beyond the defaults, and symbols are things you say out loud.
- Choose Apple menu → System Settings, click Keyboard in the sidebar, and go to Dictation.
- Turn Dictation on, and check that auto-punctuation is on if you want commas, periods, and question marks inserted for you.
- Start dictating, and say "new line" or "new paragraph" where you want a break.
- Say "numeral" before a number to format the next phrase as a number.
- Say "caps on" to write the words that follow in title case and "caps off" to return to normal. Say "all caps" to capitalize the next word.
- Say the name of a symbol to insert it, such as "at sign", "percent sign", "hashtag", or "dollar sign".
- Read the result back. Anything you still had to fix by hand is work a formatting pass would have to do.
Formatting pass vs inverse text normalization
Inverse text normalization is the older, narrower name for part of this job: turning spoken forms into written ones, such as "ten thirty" into 10:30. The Whisper paper notes that many speech recognition systems include some level of it, and that it is often simple or rule-based and detectable from what it leaves unhandled, such as never including commas.
Newer open models moved some of that work into recognition itself. Whisper was trained to predict the text of transcripts as written, which its authors say removes the need for a separate inverse text normalization step. NVIDIA describes Parakeet TDT 0.6B v2 as supporting punctuation and capitalization directly.
A formatting pass is everything that still happens after that. Punctuation from the recognizer is a starting point. Removing fillers, applying your own terms, and matching the destination app are separate steps that a recognizer does not perform.
How to add your own formatting rules in macOS
macOS has a small rule-based formatting layer of its own, which is a useful picture of what a deterministic rule is: the same input always gives the same output. Text replacements swap one string for another, and smart quotes and dashes change punctuation as text is entered.
Apple notes that text replacements, smart quotes and dashes, and smart links are not available in all apps. If text arrives with curly quotes or a substitution you did not expect, check these settings before you look anywhere else.
- Choose Apple menu → System Settings, then click Keyboard in the sidebar.
- Click Text Replacements.
- Click the Add button, type the text to replace and its replacement, then click Add.
- To change quote and dash behavior, go to Text Input in Keyboard settings and click Edit.
- Turn "Use smart quotes and dashes" on or off, choose quote styles if you want them, then click Done.
- In an individual app, choose Edit → Substitutions to turn Text Replacement on or off for that app.
Questions people ask.
What is post-processing in dictation software?
Everything that happens to the transcript after the speech model produces it: punctuation, capitalization, number formatting, filler removal, spoken commands, and tone. It is the stage that turns a transcript into writing, and the stage where tools differ most.
Does the formatting use AI?
In most current tools, partly. A small language model handles judgment calls like tone and sentence flow. The better tools handle anything with one right answer, such as times and punctuation, by fixed rule instead, because a model is only usually right.
Can I turn formatting off?
A good tool lets you choose the level, from verbatim to fully polished, and switches to verbatim automatically in terminals and code editors. Less formatting also means faster text, which is a reasonable trade for a quick reply.
Does Apple Dictation add punctuation automatically?
Yes, in supported languages. Apple’s Dictation automatically inserts commas, periods, and question marks, and you can turn this off in System Settings under Keyboard, in the Dictation section. Other punctuation is entered by saying its name.
What is inverse text normalization?
It is the step that turns spoken forms into written forms, such as spoken numbers into digits. It is one part of a formatting pass, and some newer speech models produce punctuation and capitalization directly so that less of it is needed.
How do I make dictation write numbers as digits?
In Apple’s Dictation, say "numeral" before the number, which formats the next phrase as a number. Tools with a fuller formatting pass convert numbers, times, and prices without a command.
