LiblibAI Launches "Xingliu Agent": Building on Lovart's Global Viral Success to Define China's Design Agent

On July 3, Gaorong Ventures portfolio company LiblibAI announced the Chinese official version of Lovart: "Xingliu Agent." Built on Lovart's full-stack capabilities spanning "generation—editing—layout—delivery," the product has been deeply optimized for Chinese semantics, traditional Chinese aesthetics, and local usage scenarios, aiming to provide Chinese creators with an industry-leading one-stop AI design experience.

Try Xingliu Agent now: https://www.xingliu.art/

Lovart: The Globally Viral AI Design Agent

In May 2025, Lovart's Beta version launched worldwide, attracting over 100,000 creators from more than 70 countries to its waitlist in just five days — becoming a global breakout hit in the Agent category. As a design Agent with capabilities spanning "generation + multi-turn dialogue + editable canvas + multimodal output," Lovart has been dubbed "the next-generation AI design tool."

Lovart's design made it onto X's trending list, earned a like from Elon Musk, and quickly generated tens of thousands of replies and millions of views.

Compared to traditional AI image generation tools, Lovart offers an experience much closer to real design workflows: users interact with the Agent through natural language dialogue, pitching ideas, iterating repeatedly, previewing results in real time, and ultimately completing exportable finished works.

Xingliu Agent: A One-Stop AI Design Platform That Truly Understands Chinese Creators

As Lovart's official extension for the China market, "Xingliu Agent" is far from a simple interface translation — it represents a comprehensive localization effort spanning language models, workflow architecture, and aesthetic styles. Building on Lovart's mature framework, the LiblibAI team retrained an Agent model adapted for Chinese semantics, constructed a dedicated design task pipeline system, and built in a visual style library covering Chinese cultural contexts — truly achieving an AI creative design Agent "tailor-made for Chinese creators."

From creative conception to design delivery, Xingliu Agent can execute the entire process automatically. Based on a single sentence or set of inputs from the user, the system automatically breaks down tasks, generates images, makes localized edits, arranges layouts, constructs videos, and ultimately exports high-quality visual works.

Beyond excelling at graphic visual design, Xingliu Agent has also fully integrated multiple advanced video generation models (such as Kling, Hailuo) and 3D model generation modules (Tripo, Hunyuan), enabling one-stop completion of the full video creation pipeline from script writing, storyboard planning, to character motion generation, lip-sync, and post-production export. In addition to script breakdown, character lip-sync, motion generation, and video layout, Xingliu can automatically add voiceover, music, and subtitles, outputting short-form video content over one minute in length, optimized for real social media platforms.

Xingliu Agent-generated one-minute video

In 3D scenarios, Xingliu supports rapid character + scene setup, with generated results ready for reference modeling, interactive preview, or digital human training material.

Xingliu Agent-generated 3D model

This cross-modal closed loop from "image → video → 3D" dramatically expands the boundaries of AI design, transforming creators from single-task executors into true "multi-format content directors."

Vertical Agents: From "Tools" to "Delivered Results"

As the AIGC industry rapidly evolves, vertical Agents are becoming the next core competitive battleground. Through an "automatic generation + deep canvas interaction" architecture, Xingliu Agent tightly couples large model generation capabilities with creative workflows: on one hand, it semantically drives automatic task breakdown and delivery of high-quality first drafts; on the other, through infinite canvas and tab-level editing, it supports designers in fine-grained collaboration with AI within a visual environment.

This mechanism achieves a closed-loop full pipeline of "generation—editing—arrangement—delivery," freeing creators from the constraints of single-asset output and enabling rapid switching between roles, scenes, and media within a unified space. The launch of Xingliu Agent marks the moment China's AI design begins moving from "task-based tools" to the "Agent" era, opening up new efficiencies and possibilities for future brand creativity, digital content, and virtual asset production.

As large models enter deeper waters, vertical Agents are becoming the most commercially valuable form of AI application. Compared to general-purpose tools, they feature clear semantic scenarios, reusable task pipelines, and specialized feedback loops — with the potential to truly achieve an intelligent service loop that "understands the industry, executes effectively, and delivers results."