AI Picasso

The core feature turned a written prompt into AI illustration output, with style presets so you didn't have to phrase things like a professional prompt writer. It ran on Stable Diffusion, which is why it could reach a broad audience with no design background. A partnership with the Japanese illustration library Irasutoya even generated images in that recognizable style.

You uploaded roughly 10 to 20 photos of yourself, and the app trained a personal model to produce avatars in looks you'd never appear in otherwise. This was the feature the app was best known for, and it drove most of its social sharing. The trade-off was that results depended heavily on photo quality and variety.

Generated avatars could be turned into custom LINE sticker packs, then sent inside the messaging app. It tied the product to a platform Japanese users already lived in, rather than asking them to share images elsewhere. The catch is that the sticker route only made sense for people on LINE.

Released in January 2024, AI Dance generated a short dance video from a single photo, and the company called it the first of its kind in a consumer mobile app. You uploaded one image, ideally a full-body shot, and the tool produced a clip that matched a dance pose sequence. The output was built for TikTok and Shorts, where vertical video performs best.

Beyond people, the app could stylize pet photos into cute reimagined versions and generate profile images that played up specific traits. These smaller features rounded out the avatar lineup and gave users a reason to keep generating. They were secondary to avatars but useful for variety.

AI Picasso offered an API around its own generation tech. That lineup took in an anime-focused model called Picasso Diffusion, hosted on Hugging Face, plus a beauty model tuned for Japanese faces. The company pitched these at businesses that build ads, magazine art, and manga. It also released open image models under the Emi name for commercial use.