# Keyboard Crayons

> Makes pictures and signs out of plain keyboard characters (ASCII art) that paste anywhere text goes:
> a README, a code comment, a chat, a terminal, a plain-text email or a printed note. Free in the
> browser: big block letters from the page's own five-row font, six frames, five small creatures with
> a speech bubble, a photo turned into text, and a sheet of paper to stamp pieces onto and paint, erase
> or type on one character at a time. A metered run draws a small picture of anything you name inside
> a box you choose. Derived from @nousresearch/ascii-art.

https://keyboard-crayons.skillsafe.ai/

## What it does

- Big letters: A to Z, 0 to 9 and common punctuation in a five-row block font, made of #, @, *, o or $, optionally extra wide; words wrap to the paper width.
- A creature says: a cat, owl, robot, dog or fish under a speech bubble that wraps the words.
- A photo as text: a dropped picture is read in the browser and turned into characters by brightness (16 to 80 wide, with a version for dark backgrounds). The picture is never uploaded.
- The paper: 40, 60, 72 or 80 characters wide; pieces are stamped one under another at the left, middle or right; paint and erase by dragging, type at a cursor, edit as text, undo.
- Take it with you: a frame (line, hash, stars, rounded, dots), then copy or download as plain text, a Markdown code block, // or # comments, a /* */ block or an HTML comment (with a warning when the picture would break the comment), a PNG of the paper, or print.
- Saved sheets (opt-in) keep the paper, the frame and the latest sketch in the account.

## The paid task

One task, `sketch`, on gpt-terra, metered per run. The input is six string fields: `task`, `subject` (what to draw), `with` (up to three more things, or none), `box` (columns x rows), `style` (line or shaded) and `facts` (the page's decisions). The page decides before the run: a box up to 8 rows tall fits one thing, up to 14 rows two, taller three; things past that are left out, and words in quotes are never drawn into the picture. The verdict is simplified when anything is left out, otherwise as-asked.

The reply is plain text in sections: FIT, TITLE, ART (the picture, ending at END ART), CAPTION, LEFT OUT and KEY. The page re-checks it: rows wider or longer than the box are cut, characters outside printable ASCII become ?, and any difference between what the page left out and what the reply left out is listed.

- API tutorial: https://keyboard-crayons.skillsafe.ai/api.html
- Source skill: https://skillsafe.ai/skill/@nousresearch/ascii-art (nousresearch/hermes-agent)