Comparison of video generation models between China and the US, China is currently leading
I watched a review on YouTube by the channel Dom the AI Tutor | Tech Tutor Zones uploaded on February 11, 2026, evaluating four well-known video generation models,
the selected ones are Kling 3.0, Seedance 2, Veo 3.1, and Sora 2.
Veo 3.1 was released by Google on October 16, 2025,
Sora 2 was released by the American company OpenAI on September 30, 2025 (US time),
Kling 3.0 (Kling 3.0) was released by Kuaishou on February 5, 2026,
Seedance 2.0 was released by ByteDance's Ji Meng AI on February 7, 2026.
It just happens to be two Chinese and two American models.
The evaluation was conducted by inputting the same prompt to let the four models generate videos separately, with a total of twenty scenes, including high-speed drone first-person perspective, actors speaking, police pursuit, human vs robot combat, game combat scenes, beautiful women discovering that the pilot has disappeared and is flying a private jet, etc. The account did not state which one had the best effect, but directly uploaded the videos generated by each model for every scene, allowing viewers to make their own judgments.
After watching the first few videos, I already had the answer, believing that the effect was Seedance 2.0 > Kling 3.0 > Sora 2 > Veo 3.1, and the more I watched, the more I felt this way. Seedance clearly was the best, with very good sound, visuals, and logic. However, the Veo 3.1 developed by Google performed the worst, with many obvious errors in many scenes.
Then I checked hundreds of English comments, and the ranking was surprisingly consistent, reaching 100% consistency,
meaning the effect was Seedance 2.0 > Kling 3.0 > Sora 2 > Veo 3.1.
It is worth noting that in this competition of the four models, the top two in terms of effect were both Chinese models.
Original article: toutiao.com/article/1857029983792259/
Statement: This article represents the personal views of the author.