Image to ASCII Art

about this tool

How a picture becomes characters

The picture is cut into a grid with one cell per character: as many columns as you ask for, and half as many rows per column as a square grid would have, because a character in a monospace font is about twice as tall as it is wide. Each cell's darkness is averaged over its pixels and replaced by the character with about that much ink, from a space for blank paper to the densest character in the set, such as @. Transparent pixels count as paper, so a logo with no background comes out with nothing around it.

The settings

• Width is the number of characters per line, from 8 to 400. Around 80 fits a terminal or a code comment; 150 or more shows detail but needs a wide window.
• Characters is the set used for the shades. Standard has ten, from .:-=+*#%@. Detailed has seventy, from Paul Bourke's well-known ramp, for finer shading. Blocks uses the shade characters ░▒▓█ for a solid, pixel-like look. Custom takes your own, lightest first; each must be one column wide, so emoji and East Asian characters are refused.
• For a dark background gives the bright parts the dense characters, for light text on a dark terminal. The preview turns dark to match.
• Dither passes each cell's rounding error to its neighbours, the Floyd–Steinberg way, so a short set shows gradients as a mix of characters rather than bands.
• Stretch contrast maps the picture's darkest and lightest parts (the 2nd and 98th percentiles, so a stray pixel does not count) to the two ends of the set. Turn it off to keep the picture's own brightness.

Using the result

Copy it, or save it as a .txt file named after the picture. Transparent margins are cut off before the picture is measured, so a logo with no background starts at its own edge, and blank lines above and below and spaces at the ends of lines are left out; they are invisible and only get in the way. Paste it somewhere with a monospace font, such as a code block, a terminal or a README, or the columns will not line up.

Limits

A picture larger than 1200 pixels on its longer side is scaled down to that before it is measured, which is more than 400 columns ever need; an SVG is drawn at that size, being a vector, so a small icon stays sharp. A very tall, narrow picture is drawn narrower than asked, so it keeps its shape in at most 600 lines; a thin enough strip can come out narrower than 8 characters. Each pixel counts by its lightness alone, ignoring hue, so a red and a green of the same lightness become the same character. Stretch contrast multiplies differences by at most four, so a page that is nearly all one shade stays blank rather than turning faint noise into ink. Nothing is uploaded.

A busy background is the most common reason art comes out muddy; remove the background first, or raise the contrast with image filters.

questions

Why does my ASCII art look stretched or squashed?

A character in a monospace font is about twice as tall as it is wide, so the converter uses half as many rows as a square grid would. Pasted somewhere that is not monospace, the characters have different widths, so the columns drift and the picture distorts. Paste it into a code block, a terminal or a plain-text file shown in a monospace font. Extra line spacing has the opposite effect, stretching the picture taller.

What does “for a dark background” do?

It swaps which end of the picture gets the dense characters. On paper or a light editor, dark parts of the image become @ and # and light parts become spaces. In a terminal with light text on black, that comes out as a negative, so this option gives the bright parts the dense characters instead.

What is dithering for?

A short character set has only a few shades, so a smooth gradient turns into bands. Dithering passes each cell’s rounding error on to its neighbours (the Floyd–Steinberg method), so the shades in between appear as a mix of the two nearest characters. It helps most with the blocks set and short custom sets.

Which pictures work best?

Ones with a clear subject and strong contrast: a logo, a portrait against a plain background, a silhouette. Fine texture is lost at a hundred characters wide. Transparent areas stay blank, so a logo with its background removed comes out cleanly; the stretch contrast option spreads a dull photo across more of the characters, up to four times its own contrast.

Is my image uploaded?

No. The image is decoded and converted in your browser, and the text is made in the page. Nothing is sent anywhere and nothing is kept.