GPT-56T 861 —
MUSE-SPK 837 +0.2%
GPT-56SC 790 -4.6%
GLM-5 781 -0.4%
CL-OP55X 780 -5.1%
GROK-46H 780 -5.1%
QWEN-38X 748 -9.2%
GPT-6A 743 -9.4%
KIMI-K3X 742 —
CL-FAB5H 698 -6.1%
CL-OP5H 675 -6.2%
GEM-38FH 672 -0.7%
CL-OP5X 670 -5.5%
CL-OP55H 668 —
CL-OP46H 657 -5.9%
CL-OP47H 648 -6.1%
GPT-56S 618 -0.6%
GEM-37FH 610 -7.2%
GEM-36FH 593 —
CL-OP48H 588 —
CL-OP47 581 -0.2%
GEM-35FH 580 —
GPT-55H 541 -7%
INKL 531 —
GEM-31P 512 -0.2%
CL-OP46 498 +0.4%
GEM-3P 498 -0.2%
CL-OP48 492 +0.4%
GPT-52 464 —
GPT-55 423 —
GPT-56T 861 —
MUSE-SPK 837 +0.2%
GPT-56SC 790 -4.6%
GLM-5 781 -0.4%
CL-OP55X 780 -5.1%
GROK-46H 780 -5.1%
QWEN-38X 748 -9.2%
GPT-6A 743 -9.4%
KIMI-K3X 742 —
CL-FAB5H 698 -6.1%
CL-OP5H 675 -6.2%
GEM-38FH 672 -0.7%
CL-OP5X 670 -5.5%
CL-OP55H 668 —
CL-OP46H 657 -5.9%
CL-OP47H 648 -6.1%
GPT-56S 618 -0.6%
GEM-37FH 610 -7.2%
GEM-36FH 593 —
CL-OP48H 588 —
CL-OP47 581 -0.2%
GEM-35FH 580 —
GPT-55H 541 -7%
INKL 531 —
GEM-31P 512 -0.2%
CL-OP46 498 +0.4%
GEM-3P 498 -0.2%
CL-OP48 492 +0.4%
GPT-52 464 —
GPT-55 423 —
← Back to feed

SeedDance 2.0 Holds Both Arena Video #1 Spots at Elo 1450 — 79 Points Ahead of Google Veo 3.1

ByteDance’s Dreamina SeedDance 2.0 was added to the Arena Video Edit leaderboard on April 21. A concurrent arxiv paper confirms its performance across the broader Arena video evaluation surface: #1 on Text-to-Video with Elo 1450 (±15), and #1 on Image-to-Video with Elo 1449 (±11). On T2V it leads Google’s Veo 3.1 audio 1080p by 79 Elo points; on I2V it leads Grok Imagine Video 720p by 29 points.

The result stands out on resolution grounds. SeedDance 2.0 achieves the top T2V and I2V rankings at 720p, outperforming models that generate at 1080p. Rank Spread of 1↔1 on both leaderboards indicates the model holds its position consistently across evaluation dimensions, not just in average voting.

Key Numbers

  • Arena T2V Elo: 1450 (±15), rank #1
  • Arena I2V Elo: 1449 (±11), rank #1
  • T2V lead over Veo 3.1 audio 1080p: 79 Elo points
  • I2V lead over Grok Imagine Video 720p: 29 Elo points
  • Output resolution: 720p (beats 1080p competitors in human preference voting)

Architecture

SeedDance 2.0 is built on a unified multimodal audio-video joint generation architecture. Text, image, audio, and video inputs are processed together rather than fused post-hoc. Native audio generation is part of the same model — not a downstream pipeline. The result is audio-visual synchronisation that originates from the generation step rather than being imposed on it after the fact.

The model accepts up to 20-second clip generation and supports reference-controlled creation, video continuation, and editing tasks within a single system. Its predecessor, Seedance 1.5 Pro, required separate audio handling.

Context

SeedDance 2.0 launched publicly in February 2026. The arxiv paper capturing the Arena leaderboard data accessed results on April 8; Arena formally added the model to its Video Edit category on April 21. ByteDance now has confirmed top rankings across three separate video evaluation surfaces: SeedVideoBench 2.0, Arena T2V, and Arena I2V.

ByteDance has two major AI products in production at comparable performance tiers simultaneously: Seeduplex for full-duplex voice and SeedDance 2.0 for video generation. On video, the 79-point Arena lead over Google Veo 3.1 is the widest gap any model holds over its nearest competitor in any Arena vertical tracked by Stack Futures.

The practical division of turf: Happy Horse 1.0 leads in pure visual aesthetics on some subcategory votes; SeedDance 2.0 leads when native audio, multi-shot narrative, or reference-image fidelity is the evaluation criterion.