Text tools, grouped by job

Orientation

The two kinds of text tools in this app

Everything here is one of two things, and knowing which changes how you use it. 83 routes run inside your browser: no upload, no cap, instant. 345 routes send your text to our server so a language model can read it, capped at 5 runs per rolling 24 hours and 12,000 characters a run.

A further 38 routes handle image and video watermarks, where a file genuinely is uploaded to our media workflow, and 15 are reference pages with no tool on them. 483 in total. Other directories of text tools do not need this distinction because everything they host is client-side. Ours is mixed, so it is worth five seconds of your attention before you paste something.

Sequence

The order to use the text tools in

Most jobs that start in a chat window and end in a published document pass through several tools. Running them in the wrong order wastes AI runs on problems a free tool would have removed.

First, clean locally

Strip invisible characters and normalize whitespace. Free, unmetered, nothing uploaded, and it stops a metered run being spent on debris from the clipboard.

Second, convert structure

Case, JSON, Markdown, HTML stripping, splitting, find and replace. Also local, also free. Anything a fixed rule can do should be done before a model is involved.

Third, spend a run

Rewriting, generating or analysis. This is the only step that costs you one of five, so it should happen once, on text that is already clean and correctly shaped.

Last, scan again

Anything that passes back through a rich text field can pick up fresh invisible characters. A second local scan costs nothing.

Workflow

How to pick a text tool in three steps

One question answers it almost every time.

Step 1

Ask whether a fixed rule could do the job. Removing characters, converting case, reformatting JSON and counting words all have exactly one right answer, so no model is needed.

Step 2

If the answer is yes, take the local route. It is free, instant, unmetered and your text stays on your device.

Step 3

If the answer is no, match the tool name to the task as closely as you can. The model's instruction is built from the route name, so the specific tool outperforms the generic one.

Free and unmetered

The 83 local text tools, by job

Twenty-four distinct operations, several with settings rather than one fixed behaviour.

Remove

Hidden characters, accents and diacritics, spaces, line breaks, duplicate lines, HTML tags, Markdown syntax, and em dashes. There is also a replacer that swaps em dashes for a string you choose.

Convert

Case, with Title, UPPERCASE and lowercase. Unicode normalization across NFC, NFD, NFKC and NFKD. URL encoding and decoding. Pixels to rem and back, with a root font size you can set.

Reshape

JSON minifying and prettifying, delimited text to a JSON array including a custom delimiter, Markdown formatting, text splitting on a delimiter, and find and replace.

Measure

Word counting, token counting, and a scanner that reports 14 invisible characters with their codepoints alongside 17 AI writing patterns.

Tidy

A general cleaner with switches for normalizing spaces, collapsing blank lines and straightening quotes and dashes, so you can run several passes in one go.

Costs a run

The 345 text tools that need a model

160 rewriters, 51 generators and 134 checkers and analyzers. The three behave differently enough that picking the wrong category is the usual reason a run disappoints.

A rewriter needs finished text and keeps your meaning. A generator needs a brief and writes something new in the format the tool name implies, so three words in gets three words of substance out. An analyzer reads and reports without changing anything, and is instructed never to claim certainty about authorship or which model wrote a text, only to describe observable signals and their limits.

Different rules

The 38 image and video routes

These handle watermarks in files rather than characters in text, so the rules that apply everywhere else do not apply here. A file is uploaded to our media workflow, because that is the only way the check or the edit can run.

That also means the character limit is meaningless for them and the local-processing promise does not cover them. If your job is taking a logo out of a picture, that is where to go, and it is not one of the 83.

Plan the day

Making five runs last

Five per rolling 24 hours across all 345 AI tools, not five per tool, with the window rolling forward from each use so they return individually rather than all at midnight.

Do the free work first

Every local pass you run before a metered one improves what the model receives and costs nothing.

Trim to the passage that matters

The ceiling is 12,000 characters, but a focused excerpt usually produces a better result than a whole document.

Write a real brief for generators

A vague brief means a second run, and the second run is the expensive one.

Do not re-check after every edit

Collect the changes, make them all, then spend one run on the result.

Straight answer

Are all the text tools here private?

No, and the honest answer has two parts. The 83 local tools are private by construction: nothing is transmitted, so there is nothing to log or lose. That is not a policy, it is how they work.

The 345 AI tools send your text to our server, because a model cannot read it otherwise. Nothing is written to storage and the only record kept is a run counter with a number and an expiry, but the text does travel. Directories that promise your data never leaves your device can say that because everything they host is client-side. We will not borrow the line for a catalogue where it would be false two thirds of the time.