Hello Keyboard Text Compare: Check Your Revision

Nobody thinks about a text compare tool until the wrong version goes out. A “not” that vanished. A figure that moved. A qualifier that got lost somewhere between draft two and draft three. Hello Keyboard Text Compare is a free browser utility for exactly that moment: you paste the original passage and the revised one, pick words, characters or lines, and see what actually changed before you publish or send.

There is no account and nothing to install. Open the page, paste two blocks of text, press one button. That is the entire product, and for this job that is a strength.

Why a revision checker earns its place

Consider how copy gets edited now. A draft goes out, comes back marked up, passes through a rewriting tool, then comes back again. By the third round, nobody can quote what version one said. The new text reads well, so it gets approved. Reading well and saying the same thing are two different tests.

This is where a visible diff beats memory. Holding two versions in your head is not a comparison; it is an impression. Your eye skips familiar phrases and your brain fills in what it expects. A highlighted diff fills in nothing. It marks the inserted word, the deleted clause and the paragraph that did not move.

The awkward truth is that most damage in an edited passage is small: a dropped condition, a softened claim, a name spelled differently, an instruction that lost its “unless”. Rereading tends to miss these, because a familiar text gets skimmed rather than checked.

How the text compare tool works

The page has two fields: original text and changed text. You choose a comparison mode and select Compare text. The result marks removals and additions, keeps the unchanged context around them, and adds a summary of how many units changed or stayed the same.

The calculation runs in your browser, so JavaScript must be enabled. The tool page says the comparison runs in the page and that pasted text stays in the current browser session. Under the hood, the site’s open-source credits page states that text comparison uses jsdiff, a widely used JavaScript diff library based on the Myers difference algorithm.

It is worth being precise about what “in the browser” covers. Client-side processing means the task runs on your device instead of a remote server, so the input does not need to leave it, as ByteQuant explains. It does not mean the page makes no network requests at all, and it is not a shortcut to legal compliance. The section on limits below spells out what this particular site does load.

Word mode: the readable default

Word mode is the right starting point for prose. It counts words as units and keeps the spacing around edits visible, so an inserted term reads as a term. Articles, emails, instructions, policy text, anything written in sentences: start here. It is also the least noisy mode, because it reports what moved at the level people actually read.

Character mode: for the tiny edits

Character mode exposes the smallest changes: a spelling fix, a swapped punctuation mark, a capital letter, a hyphen that appeared or vanished. Use it for proofreading passes, or when you suspect something changed but cannot spot it.

On small edits it is precise and quiet. It gets noisy when whole words were replaced, because the algorithm matches shared letters inside different words and the result fragments. If a passage was substantially rewritten, go back to word mode.

Line mode: when each line is a unit

Line mode groups additions and removals by complete lines. It suits notes, snippets, configuration text and changelogs, where a line is the meaningful unit. Paste two versions of a config block and line mode shows which lines left and which arrived, without word-level clutter. Note that a one-character edit marks the whole line as removed and re-added.

The rule of thumb: words for writing, characters for proofreading, lines for structured text.

Swap, clear, and keeping both inputs

Swap exchanges the two inputs, which reverses the perspective: additions become removals and vice versa. It helps when you want to read the edit from the other side, or when you pasted the texts in the wrong order.

Clear empties the fields for a new comparison. Neither control edits your text, and both passages stay available while the result is on screen. That matters more than it sounds: you can read the diff, return to the original sentence and check the surrounding paragraph without losing your place.

What the change summary actually tells you

The summary counts added, removed and unchanged units for the mode you selected: words, characters or lines. It measures textual difference and nothing else.

One detail catches people out. In word mode, spacing stays visible around edits, but a space is not counted as a word. You may see a gap change while the word count stays put. That is not a bug; it tells you the change is not a word-level change.

The tool page gives its own example. Original: “Send the report today.” Revised: “Send the final report today.” The result reports one word added, “final”, highlights it, and keeps the rest of the sentence visible around it. Exact unchanged counts can depend on how punctuation such as the final full stop is tokenised, so run the example yourself if you need the precise numbers.

Notice what the summary does not say. It reports that a word appeared, not what the word does to the meaning. Here it does a lot: “final” turns a routine send into a last call. The tool found the edit. Deciding what the edit means is your job.

Practical workflows for comparing two texts

Reviewing AI-rewritten copy against the original

This is the strongest use case, and the one most often skipped. You have a paragraph you wrote and a version a model rewrote. The rewrite reads better. Does it still say the same thing?

Paste both, run word mode, and go through the additions and removals. Look for specific things: a new qualifier that weakens a claim, a limitation dropped because it made the sentence clunky, a number that shifted, a changed name, a condition that disappeared. Rewriting models often smooth text, and smoothing is where meaning slips. They can also add details that were never in the source, so treat every addition as a claim to verify.

The diff will not tell you whether the new version is accurate. It tells you where to look, turning a vague unease into a short list of places to check.

Checking a document revision before sending

Contracts, briefs, instructions: anything with a recipient and consequences. A revised version comes back with tracked changes off, or as a clean copy from someone else’s editor. Paste the version you sent and the version you received.

Use word mode for the prose. Switch to character mode where a clause looks identical but feels different; usually a comma or a modal verb moved. Use line mode for lists, steps or numbered conditions, because a deleted step is easier to spot as a missing line.

Then do the part the tool cannot. Read each changed passage in full, in context, and ask whether the revision still means what was agreed. A diff is an index, not a verdict.

Copy edits, messages and configuration snippets

The tool page lists these as everyday uses. A copy-edit pass on a landing page. A message you rewrote twice. A config block where one value changed and you need to confirm nothing else did.

For the config block, use line mode. For the message, word mode. For the copy edit, word mode first, then character mode on the headline alone, since headlines are where stray punctuation hides.

Comparing multi-line text without losing the structure

Both fields accept paragraphs and multiple lines. Choose Lines and the result is grouped by complete lines, which keeps the document’s shape visible instead of collapsing it into a run of inline marks.

This is where a browser diff beats two documents open in separate tabs. With tabs, you are the comparison engine, and people are poor comparison engines.

Where this tool stops

The scope is narrow, and that is fine as long as you read it honestly.

It is not an AI writer, a grammar checker, a spelling checker or a fact-checker. It does not judge whether the revision preserves your meaning, whether a claim is true, whether a citation still holds, or whether the change improves the writing. It shows textual difference. Everything after that is yours.

The tool page states no maximum input size, offers no export format and mentions no saving or history. The site’s privacy notice confirms that browser-local tools create no cloud history or account and that their inputs are not saved to local storage. For developers who need diffing inside their own code, Python’s difflib compares sequences of any hashable items, most often lines of text, and can output unified, context or HTML diffs. Hello Keyboard covers the everyday version: paste, compare, read, close the tab.

Client-side tools also carry a performance trade-off: they must download their JavaScript before doing anything, and large bundles slow the first load, as ToolBrew notes. A text comparison utility is light enough that this rarely matters, but it is the price of keeping the work on your machine.

What “in-browser” does and does not mean here

In-browser processing is a real architectural difference from uploading text to a server. It is not the same as a page with no third-party requests, and in this case the privacy notice is specific:

The site uses Google Analytics 4 to measure visits, referrers, device categories and interactions. Google may set cookies or similar identifiers.
All Hello Keyboard sites load Google AdSense Auto ads, whose partners may use cookies to serve, measure and personalise ads.
The site is hosted on Cloudflare, which receives connection data such as your IP address, the requested URL and request headers.
The notice states that tool inputs, files, filenames, results and clipboard content are not included in analytics events.

That last statement is the site’s own claim. It is consistent with the architecture, but it is a promise you are trusting, not something you have verified. For confidential text, check the network panel in your browser’s developer tools while you run a comparison, or use non-sensitive sample text when evaluating this or any other browser tool. Browser extensions, clipboard history and shared computers remain outside the site’s control either way.

The habit worth building

The tool is the easy part; the habit is the hard part. Most people skip comparing versions because it feels tedious, and tedious steps get skipped until something embarrassing ships.

So build the smallest version of the habit. Before you publish an edited page, paste both versions into a text compare tool and spend 30 seconds on the additions. Do the same before sending a revised document. Always do it before accepting an AI rewrite, because that is the case where the text reads better and may mean less.

What you gain is not accuracy but attention, aimed at the right sentence. The tool finds the difference. Whether the difference is a mistake is a call no browser utility makes for you.

Free, no account, three modes, one button. For a job this small, that is enough. The rest is reading the highlighted word and asking: did I mean to change that?

Additional resources

This article includes content created with AI.

Hello Keyboard Text Compare: Check a Revision