Seedance 2.5 strengthens the competitiveness of China's AI video models, but has limited short-term impact on Kling

Overview

This report evaluates Seedance 2.5, disclosed by Bytedance at the Volcano Engine Force conference, with a focus on its impact on the competitive landscape and commercialization process of China's AI video generation model market. Citi believes the core value of Seedance 2.5 lies in improving video length, consistency, and controllability, thereby enhancing creators' workflows, but the upgrade is more of an efficiency and user-experience enhancement than a fundamental technological breakthrough.

Core views

The core views include: first, with 30-second native video generation, 50 multimodal reference inputs, and localized re-editing capabilities, Seedance 2.5 is expected to further strengthen Seedance's leading position in the Chinese market; second, these features can improve multi-shot visual consistency, character-detail fidelity, and controllability in video generation, enhancing practical value in professional production scenarios; third, Kuaishou's Kling faces limited short-term impact because its revenue mainly comes from overseas markets, while Kling may also quickly introduce upgrades in video length, controllability, and consistency; fourth, other players in the industry are still catching up, and the competitive landscape is not yet settled.

Analysis framework

The report uses product feature comparison and competitive landscape analysis, comparing Seedance 2.5's new features with peer models' typical video duration, reference input capabilities, and editing workflows, and further assesses the impact of these capabilities on user efficiency, commercialization potential, and competing products such as Kling.

Methodology notes

• Competitive landscape analysisFeature upgrade comparison

  Assess model competitiveness across three dimensions: video length, number of reference inputs, and editing controllability.

  The report compares Seedance 2.5's 30-second video generation with peers' typical single-segment generation of 15 to 20 seconds, and highlights the improvements in consistency and professional production efficiency from multimodal reference inputs and localized editing.

• Commercialization analysisUser efficiency and production workflow

  Improvements in model capability unlock commercialization value by reducing stitching, regeneration, and manual correction.

  Longer videos, stronger consistency, and localized editing capabilities can lower post-production costs for professional creators and improve the usability of AI video tools in commercial production scenarios.

• Regional market exposure analysisDifferences between China and overseas markets

  The regional composition of revenue and users across products affects how competitive impact is transmitted.

  The report believes Seedance will strengthen its presence in the Chinese market, but since most of Kling's revenue comes from overseas, the Seedance upgrade has limited short-term impact on Kling.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Bytedance / Seedance 2.5

  Core subject of analysis in the report.

  Strengths

  Strong motion smoothness and visual consistency, with the new version adding 30-second video, 50 multimodal reference inputs, and locally controllable editing.

  Weaknesses

  The report believes the upgrade is not a breakthrough change, and its overseas expansion is not aggressive.

  Comparison

  Seedance 2.0 was already in a leading position in the video generation model industry, and Seedance 2.5 goes a step further than most peers in video length and controllability.

  Risks

  If actual release performance falls short of expectations, or if user-efficiency gains are limited, commercialization potential may be lower than market expectations.

• Kuaishou / Kling

  Main competitor and potentially affected player.

  Strengths

  It has a solid presence in overseas markets, with most revenue coming from overseas, so the direct short-term impact from Seedance's stronger position in China is limited.

  Weaknesses

  It needs to continue improving video length, controllability, and consistency to respond to the competitive pressure from Seedance's upgrade.

  Comparison

  Compared with Seedance's further strengthening in the Chinese market, Kling's advantage is more reflected in its overseas market exposure.

  Risks

  If the pace or magnitude of subsequent upgrades lags behind, it may face pressure in user mindshare and professional creator scenarios.

• HappyHorse 1.1

  An example of an industry catch-up player.

  Strengths

  The report mentions it has already launched version 1.1, showing that industry players are still iterating rapidly.

  Weaknesses

  The report does not provide specific feature advantages relative to Seedance or Kling.

  Comparison

  As a catch-up player, its presence indicates that competition in AI video models is still accelerating.

  Risks

  If its core capabilities cannot differentiate in consistency, length, and controllability, its commercialization opportunity may be constrained.

Key data

• Report date2026-06-23 09:36:01 ETDisclosed on the report cover page.
• Seedance 2.5 launch timingExpected in early July 2026The report says this version is nearing the end of testing.
• Native video generation length30 secondsThe report says most peer models can typically generate only 15 to 20 seconds per segment.
• Multimodal reference inputsUp to 50Reference inputs include images, audio clips, and concept art, among others.
• Editing capabilityLocally controllable re-editingUsers can choose to modify areas such as the background, product, or character while keeping other parts unchanged.
• Report length8 pagesPage 1 contains the core views; pages 2 to 8 mainly cover the rating system and compliance disclosures.

Impact & implications

If successfully launched, Seedance 2.5 will reinforce Bytedance's leading image in China's AI video generation model market and push the industry from showcase-style capabilities toward a more controllable, professionally production-ready commercialization stage. From an investment perspective, the upgrade is more likely to increase market focus on the penetration and paid conversion of AI video tools rather than immediately reshaping the competitive landscape; competitors such as Kling still have room to remain competitive through overseas markets and subsequent feature upgrades.

Risks

• The report judges the Seedance 2.5 upgrade to be an efficiency and user-experience enhancement rather than a breakthrough technological leap.
• Competitors may quickly launch similar upgrades, weakening Seedance's leadership window.
• Seedance's overseas expansion is not aggressive, which may limit its global commercialization impact.
• If Kling rapidly upgrades in video length, controllability, and consistency, Seedance's competitive impact on it may decline further.
• AI video generation still faces execution risks such as long-video consistency, character-detail retention, localized editing quality, and adoption rates in professional production.

What to watch

• Whether Seedance 2.5 launches on schedule in early July 2026.
• Whether 30-second video generation can reduce seams, facial distortion, color shifts, and motion discontinuities in real-world use.
• Whether 50 multimodal reference inputs significantly improve consistency across multiple characters and shots.
• The extent of improvements in Kling's next upgrade in video length, controllability, and consistency.
• Seedance's pace of expansion in overseas markets, and whether Kling's overseas revenue and user growth remain resilient.