How the tools work

Browser tools

Conversions and generators run on this device. A result describes the input you gave it; browser diagnostics describe the events this browser received.

Standards and implementation

Encoding rules follow the standards and maintained libraries linked from Sources and licenses. That page records reference dates and credits.

What changes

Case conversion changes capitalization. Formatting changes layout according to the selected language. Each tool explains where a conversion can change the original text.

Stable rules

These tools do not fetch live prices or changing external records. Browser support and library behavior may change; versions are recorded with the source.

What a result cannot prove

A barcode image does not assign a retail product number. A missing keyboard event does not identify a hardware fault. Online color-vision diagnosis is not offered.

Corrections

Report the tool name, your expected result, and a small non-private example through the contact page when contact delivery is available.

Barcode and QR rules

EAN and UPC check digits are checked before the image can be saved. The final EAN-13 digit makes the weighted sum of the digits a multiple of ten. The default Code 128 example is an invented stock label, not a registered product.

Barcode borders use a minimum of twelve narrow-bar widths on each side. QR borders use at least four squares. Those limits protect clear space in the exported image; they do not certify a printed label. QR payloads are static text with no redirect service.

Counts and case

Visible characters use Intl.Segmenter grapheme clusters. UTF-8 bytes use TextEncoder. Code points use JavaScript string iteration, and UTF-16 code units use string length. Removing whitespace happens before the visible-character count is repeated.

Words are Unicode letters or numbers with internal apostrophes and hyphens. Sentence counts include unfinished final fragments. These are stated counting choices, not promises to match every platform. Acronym preservation is a mechanical rule for existing all-capital words.

Title-case suggestions implement common English MLA rules and the Chicago eighteenth-edition long-preposition change. Each word has a visible reason and correction. The converter uses a word list rather than a grammatical parser, so phrasal verbs, unfamiliar prepositions, proper names and compounds need review. Existing all-capital acronyms and internal capitals can be retained. AP and APA are not included.

Bytes and signals

Binary and Base64 text conversions use UTF-8. ASCII mode rejects non-ASCII characters. Binary output groups complete eight-bit bytes. URL component mode escapes separators that whole-URL mode retains. HTML entities are shown as text.

Morse playback uses the PARIS words-per-minute convention: the dot duration is 1.2 divided by words per minute, in seconds. Tone samples are written to a mono PCM WAV. Roman numerals use the conventional subtractive forms from 1 through 3,999.

Code formatters

JSON formatting edits token spacing without serializing numbers through JavaScript floating-point values. XML preserves text-node whitespace and rejects document types and entity declarations. YAML preserves comments when formatting and rejects unsafe integer conversion to JSON. SQL formatting does not execute queries.

Formatters work separately from the page and stop after ten seconds. Input is limited to two million characters. Tree views and line comparisons are abbreviated for readability; the copied result remains the complete formatted text. These are application limits, not language limits.

Draws, shuffles and passwords

Decimal range endpoints are scaled as exact decimal values before rounding to the requested grid. Each random choice uses Web Crypto with rejection sampling to avoid a modulo bias. Fisher–Yates shuffles the remaining entries. A duplicate line is a separate entry unless deduplication is selected. Teams are dealt from a shuffle; bingo uses distinct supplied entries.

Passwords choose characters independently from the allowed set. If selected sets are required, a candidate missing one is discarded and redrawn. Passphrases choose words independently from EFF’s long list. Random tools store neither lists nor generated secrets.

Writing and drawing tools use a curated subset of the credited EFF vocabulary. The acting tool uses original CC0 prompts. Category and challenge labels are editorial choices, not age ratings. No-repeat draws exhaust the selected pool; changing filters or resetting starts a new pool. Bingo cards are 3 by 3, 4 by 4 or 5 by 5; a free center is offered only on odd-sized cards. Entries are limited to 80 characters for print layout.

Colors and pixels

Conversions use sRGB channel values, with display rounding for HSL and HSV. Palettes rotate hues and adjust lightness while preserving the first seed color. Locked swatches retain their previous exact HEX values.

The image picker samples original canvas pixel coordinates after the browser decodes the file. Zoom changes the displayed size. Browser color management and image compression can affect the decoded color. Images are limited to ten megabytes and twenty million pixels.

Opacity is stored as the exact alpha byte divided by 255. HEX and the slider keep that byte; changing the slider selects another byte. Copied RGBA and HSL round opacity to four decimals, enough to recover the same byte. The opacity label also shows the exact byte fraction. HSL rounds its color coordinates for display. HSV lists color channels separately from the shared opacity.

Browser diagnostics

Keyboard and mouse tools report received events. Gamepad values come from the browser Gamepad API; missing devices remain unavailable. The dead-zone calculation subtracts the selected threshold and rescales the remaining range.

Speaker tones are generated locally and stop after two seconds. The display tool shows solid colors without flashing or claiming repair. Automated browser checks cannot establish a physical device’s health, audible output or color accuracy.

Resting gamepad samples cover three seconds for the first available controller. Axis ranges and raw/remapped traces describe that interval only. Pausing, changing the dead zone, disconnecting, or hiding the page clears the sample. It is not a pass/fail threshold for controller health.