Seedance 2.0 holds Artificial Analysis Video Arena #1 in both text-to-video and image-to-video — Veo 3.1, Kling 3.0, Sora 2 below
Seedance 2.0 holds the top ranking on Artificial Analysis Video Arena leaderboards in both text-to-video and image-to-video as of May 2026 — ahead of Kling 3.0, Veo 3.1, and Sora 2. The standing tells the story: a Chinese-lab model leads the Western competitor set in independent benchmarks on the same week Google launches Gemini Omni as the supposed Veo successor.
The ranking is the policy data point. Through 2024-2025 the Western frontier-lab framing treated Chinese video-generation labs as a quality tier behind OpenAI, Runway, and Google. Seedance, Kling, and Vidu are not a tier behind anymore. The Artificial Analysis Video Arena uses pairwise-comparison human evaluations from a global pool of judges — methodologically the same arena setup that gave us the text-model leaderboards — and the top of the table is dominated by Chinese labs.
The compute story underneath: Chinese video labs are training on chips they couldn't get through the H100/H200 export controls, which means they're either using domestic accelerators (Huawei Ascend 910C, Biren BR104) or finding routes around the restrictions (Singapore, Malaysia transshipment). Whatever they're using, it's working well enough to win benchmarks. If a US administration tightens controls further, the immediate competitive question is whether Western labs can keep their edge in video generation — Seedance 2.0 has already demonstrated that the edge in May 2026 doesn't exist.
TechCrunch — Google Gemini Omni turns images audio text into video → · Blockchain News — Gemini Omni Replaces Veo video model → · Medium / Analyst Uttam — Gemini Omni vs Seedance 2 →