Why run text tools locally
Text tools look simple until you paste something you would never email to a stranger: an API response with a live key, a contract redline, a donor list, a draft that is not public yet. Most “free online” counters and formatters post that paste to a server so a PHP script can run strlen() and send you HTML back. imissfiles does the opposite. Diff, word count, case, JSON, dedupe, and find-and-replace all execute in this tab. Closing the page discards the working copy. Ads sit around the workbench. They never see the string.
People arrive here from two directions. The first is a document they already converted — PDF to Text, Image to Text, or OCR PDF — and now they need to count it, recase it, or pull duplicates out of the extracted lines. The second is a developer or editor who never had a file at all: two versions of a README, a JSON blob from DevTools, a caption that has to hit 160 characters. Both jobs belong on the device that already holds the text. Uploading a JSON token to a formatter is how keys leak into access logs. Uploading a manuscript to a word counter is how drafts show up in someone else’s training set. We do not want that data and we do not take it.
A word counter is the highest-volume search in this category, so it is built as a live stats panel rather than a submit button: words, characters with and without spaces, sentences, paragraphs, lines, reading time at about 230 words per minute, speaking time at about 140, plus keyword density and the common limits for Twitter/X, meta descriptions, SMS, and LinkedIn. A five-minute speech is roughly 700 spoken words. A 500-word essay is a first-year page. Those numbers update as you type so you stop guessing.
Diff checker is the tool we spent the most time on, because a toy diff deletes a whole line when you changed one word. Ours runs a Myers line diff and then a second pass inside changed lines so a single-word edit stays a single-word edit. Side-by-side is the default on a wide screen; unified is the default on a phone. Collapse runs of more than ten identical lines or a 2,000-line config file is unreadable. Ignore whitespace, case, and blank lines when you are comparing messy exports. Similarity is 2 × unchanged lines ÷ (lines in A + lines in B), shown as a percent — useful, not a court exhibit.
JSON formatter is the privacy argument in one screen. Developers paste production payloads into beautifiers all day. Format, minify, sort keys, and expand a collapsible tree here. On failure you get the line and column, not a shrug. Case converter keeps the original so you can try Title Case, snake_case, camelCase, and CONSTANT_CASE without re-pasting. Remove duplicate lines is the list cleaner people still open Excel for: keep the first occurrence, show only dupes, or keep lines that appear once. Find and replace adds regex, whole-word, capture groups, and step-through matches.
None of these six need an account, a daily quota, or a desktop runtime. They share a shell: paste or drop a plain-text file, get a live result, copy or download. Files over about 5 MB warn instead of locking the tab; expensive work runs in a Web Worker. If your source is still a PDF or a photo, extract it with PDF to Text, Image to Text, or OCR PDF, then come back. The text never has to leave the machine it arrived on. Bookmark the hub, not a random “text converter” that stores pastes. When we add sort-lines or a slug generator later, they will use this same local shell.