OG / social image generator
Fit a background image to the 1200x630 Open Graph size and stamp a title on it, on your device.
Runs on your device. The file is never uploaded.
OG / social image generator covers the fixed 1200x630 Open Graph box with your photo, centre-crops the overflow and writes a lossless PNG with an -og suffix. A 160 pixel band across the bottom is dimmed to 45% and your title is stamped over it. That band is always drawn, even with no title, and cannot be moved or removed.
Questions
What size is the image it makes?
Exactly 1200x630 pixels, the standard Open Graph preview size, and it is not configurable. Your picture is scaled until it covers that box and then centre-cropped to it, so the proportions are kept and the overflow is trimmed. The output is a lossless PNG named with an -og.png suffix.
What is the dark strip along the bottom?
A 160 pixel band across the full width, dimmed to 45 percent of its original brightness, so a title stays readable over a busy photo. It is always applied, even when you leave the title blank, and there is no option to turn it off, move it or change how dark it gets.
Why does my title look wrong?
Because the text is drawn with a small built-in 3x5 pixel font that covers capital A to Z, the digits 0 to 9 and the space only. Lowercase is uppercased automatically and every other character, including punctuation and accents, is drawn as a blank space. The size starts at scale 8 and shrinks step by step down to 2 to fit inside the width, so a long title ends up small.
Does it write the meta tags for me?
No. It produces the image file only. You still add the og:image and twitter:image tags to your page yourself, pointing at wherever you host the PNG. It also does not fetch a page, read a headline or generate a background; you supply the background image.
Does it upload my image?
No. Decoding, the cover fit, the band, the text and PNG encoding all run in a Web Worker in this tab using the jSquash codecs. Nothing is uploaded and no font is fetched from a network. JPG, PNG and WebP inputs are read; anything else stops with "only JPG, PNG and WebP are supported".