🛰️ Daily AI Frontier
‹ back to 2026-08-05

别再吹AI生图了,不能图层编辑的AI都是“画饼”

量子位 Multimodal & Generative 思邈 2026-08-05
Representative image for 别再吹AI生图了,不能图层编辑的AI都是“画饼”

TL;DR - TuZhan Intelligence launched RabbitVis, an AI design tool built on UniWorld-Design that turns generated images into editable, reusable layered assets. It targets the post-generation work—editing, resizing, and asset reuse—that conventional image generators leave to professional design software.

  • Supports layer decomposition, transparent-background asset generation, image editing, text editing, and element-level adjustments.
  • UniWorld-Design structures backgrounds, subjects, text, and decorations as independent layers rather than flattening them into one image.
  • Reported image-to-layer results include 0.1264 RGB L1, 0.7325 Alpha Soft IoU, and a 20.43/25 VLM score.
  • The product is positioned as an end-to-end design workflow spanning generation, revision, reuse, and delivery.

view merged work →