Hey Product Hunt, avatarsniff solves a small, specific annoyance. When someone never sets a profile picture, their provider hands you a generated default: a letter on a coloured square, a flat colour block, the Gravatar silhouette, a GitHub identicon. To you it's just an image URL, indistinguishable from a real photo, so you end up showing the boring placeholder. avatarsniff reads the image and tells you it's a default, and which kind, so you can swap in your own avatar instead. What makes it hold up over time: the detection is structural, not colour-matching. It keys on shape (a flat saturated background, a near-white glyph, mirror symmetry) so it keeps working when providers add new colours. A few specifics: - Zero install dependencies. PNG/GIF/JPEG decode in the core; WEBP and SVG are opt-in so their wasm only loads if you need it. - Same API in the browser, Node, Deno, Bun, edge and workers. No framework. - One function. Pass it bytes, a URL, or canvas ImageData. There's a live demo where you can paste any avatar URL and watch it get classified: https://avatarsniff.tunc.co It's free and MIT-licensed: https://github.com/tunctn/avatar... Would love to hear where it trips up on your data. Thank you!

avatarsniff

Sniff out generic default avatars, straight from the pixels.

Sniff out generic default avatars, straight from the pixels.