ByteDance Seedance 2.0 and Alibaba HappyHorse-1.0 take top Artificial Analysis video-generation slots — Chinese labs lead text-to-video frontier as Sora exits

ByteDance's Seedance 2.0 (released February 12, 2026) and Alibaba ATH's HappyHorse-1.0 (released April 2026) now occupy the top two slots on Artificial Analysis text-to-video leaderboards. The Chinese-lab leadership in video generation comes as OpenAI deprecates its Sora product and Google's Veo 3.1 settles into the middle of the leaderboard.

Seedance 2.0's capability profile is the differentiating piece. The model supports text, image, audio, and video as inputs and generates 4-15 second clips at up to 1080p. The multi-modal input flexibility is what Veo 3.1 and Runway Gen-4.5 do not yet match. Veo 3.1 retains the audio-quality leadership — synchronized dialogue with 48kHz speech generation — but the input-modality range is now a Chinese-lab strength.

The two-model workflow becoming standard in production teams. Most 2026 production deployments now use Seedance 2.0 (or another in-editor model) for the bulk of B-roll and continuity shots, plus Runway Gen-4 or Veo 3.1 for hero shots. The single-model-for-everything pitch is no longer competitive; the differentiation is per-shot-type capability matching.

Pixflow — Best AI Video Generator in 2026 Runway Veo Seedance Kling → · FreeVideoGenerator — Best AI Video Tools 2026 Full Review Sora 2 Veo 3.1 Kling Runway →