Screenshot from Lovart official website
AsianFin -- Chinese developers are increasingly driving some of the newest artificial intelligence tools aimed at a worldwide audience.
Melvin Chen, a former ByteDance China executive, co-founded AI design startup Lovart, which officially launched Wednesday after claiming “800,000 users across 70 countries” during its beta phase.
“We will focus on North America as the first step,” Chen told CNBC in Mandarin. He previously led China operations for CapCut, ByteDance’s popular video-editing app that remains the top photo and video app in Apple’s U.S. App Store.
Lovart leverages AI to generate logos, stickers, and branding visuals from simple text prompts. Its new release includes a “ChatCanvas” feature designed to simplify specific visual edits—a task that’s difficult to explain using only words but easier with images, Chen explained.
Chen expects Lovart to surpass 1 million users within a week of launch. However, the app won’t enter China soon, as it relies on AI models from Anthropic (Claude 4) and OpenAI, which are not officially accessible in China due to regulatory restrictions and government controls blocking many foreign platforms.
While most of Lovart’s team operates from San Francisco to better localize the product, Chen said part of the production team remains in China. The startup offers a free tier alongside premium subscriptions up to $90 per month and plans to seek investor funding after scaling its user base.
Amid escalating U.S. export controls on advanced semiconductors to China, the AI race sees the U.S. leading foundational model development with OpenAI’s ChatGPT, boasting 70 million monthly U.S. users and 144.6 million in Europe. China’s AI edge, however, is increasingly seen in application development.
Chinese startups like Kuaishou’s Kling, Shengshu’s Vidu, and Manus (an AI agent platform) have gained significant global traction. Open-source Chinese AI models are popular on platforms such as Hugging Face, which shows China-developed models frequently topping usage charts.
This week, Moonshot AI’s Kimi K2 coding model ranked first on Hugging Face, followed by Alibaba’s Qwen3 and Tencent’s Hunyuan in respective categories. Open-source availability enables global developers free access to high-performance AI tools.
“China-affiliated teams are increasingly influential, driven by dense technical talent, agile development culture, and policy support for commercialization,” said Charlie Dai, VP and principal analyst at Forrester. “They excel at cost-efficient training and rapid consumer app iteration, often emphasizing open-source accessibility.”
During a recent Beijing visit, Nvidia CEO Jensen Huang highlighted the widespread adoption of Chinese-developed models like DeepSeek, noting most users run these locally worldwide. He emphasized that AI development priorities are shifting from “which model is smartest” to “which is most useful.”
Lovart exemplifies this trend by combining China-born talent with U.S.-based localization and global user focus. Its tools aim to integrate AI into creative workflows, signaling a broader shift in China’s AI ambitions—from chasing foundational model supremacy to delivering practical, globally relevant applications.
