字节跳动Seedance 2.0引爆AI视频战局

像华策影视、完美世界这些公司的股价，周一都出现了大幅上涨，中文在线甚至直接触及了百分之二十的涨停板。

Stock prices for companies like Huace Media and Perfect World saw significant gains on Monday, and COL Group even hit its twenty percent daily limit immediately.

这背后的推手，毫无疑问，就是字节跳动刚刚放出的这个“大杀器”。

The driving force behind this is, without a doubt, this "powerful weapon" just released by ByteDance.

这款名为 Seedance 2.0 的视频生成模型，虽然目前还处于预发布阶段，只在即梦AI平台上对部分用户开放测试，但它展现出的能力已经让整个科技圈和视频制作行业感到震惊。

This video generation model named Seedance 2.0, although currently still in the pre-release stage and only open for testing to select users on the Jimeng AI platform, has displayed capabilities that have already shocked the entire tech circle and video production industry.

早期的测试用户反馈说，生成的视频画面逼真得让人难以置信，几乎打破了AI内容和现实拍摄之间的界限。

Early test users reported that the generated video imagery is unbelievably realistic, almost breaking the boundary between AI content and real-life filming.

这种“电影级”的质感，不仅仅是清晰度的问题，更在于镜头运动的流畅度和视觉的一致性。

This "cinematic" quality is not just a matter of resolution, but lies more in the fluidity of camera movement and visual consistency.

这就不得不提一位来自广东的程序员王磊的体验，他只是输入了一些简单的文本提示词，就生成了一段跨越人类历史十秒钟的短片。

This brings us to the experience of Wang Lei, a programmer from Guangdong, who simply input some simple text prompts and generated a ten-second short film spanning human history.

用他的话来说，叙事流畅，甚至有一种史诗般的宏大感。

In his words, the storytelling was smooth, and it even had a sense of epic grandeur.

这得益于字节跳动背后庞大的视频数据资源，也就是我们熟悉的抖音，为模型的训练提供了得天独厚的优势。

This benefits from the massive video data resources behind ByteDance, which is the Douyin we are familiar with, providing unique advantages for the model's training.

而且，Seedance 2.0 支持多模态输入，文字、图片、音频、视频都能作为素材，这大大降低了专业视频制作的门槛。

Moreover, Seedance 2.0 supports multimodal input; text, images, audio, and video can all serve as source material, which greatly lowers the barrier for professional video production.

就连著名的游戏制作人，《黑神话：悟空》的开发商游戏科学的CEO冯骥，都对它赞不绝口，甚至评价它是“地球上最强的视频生成模型”。

Even the famous game producer, Feng Ji, CEO of Game Science, the developer of "Black Myth: Wukong," was full of praise for it, even evaluating it as the "strongest video generation model on the planet."

当然，冯骥也指出了硬币的另一面，那就是随着技术门槛的降低，人人都能低成本制作专业级内容，这势必会给传统的影视工作流带来剧烈变革。

Of course, Feng Ji also pointed out the other side of the coin, which is that as the technical barrier lowers and everyone can produce professional-grade content at low cost, this is bound to bring drastic changes to traditional film and television workflows.

同时，深伪技术带来的信任危机，也就是我们常说的 Deepfake 问题，也成为了无法回避的挑战。

At the same time, the trust crisis brought by deepfake technology, which is the Deepfake problem we often talk about, has also become an unavoidable challenge.

尽管有争议，但商业化的前景是不可否认的。

Despite the controversy, the prospects for commercialization are undeniable.

开源证券的分析师就认为，Seedance 2.0 这种能够给用户提供“导演级”精准控制的工具，会最先在AI漫画和短剧领域落地，帮助那些拥有大量IP的公司降本增效。

Analysts at Kaiyuan Securities believe that tools like Seedance 2.0, which can provide users with "director-level" precise control, will first be implemented in the fields of AI manga and short dramas, helping companies with large holdings of IP to cut costs and increase efficiency.

不过，字节跳动并不是在唱独角戏。

However, ByteDance is not performing a monodrama.

就在几天前，快手科技也发布了可灵 3.0 系列模型，同样主打多模态输入输出，瞄准的也是行业应用。

Just a few days ago, Kuaishou Technology also released the Kling 3.0 series models, also focusing on multimodal input and output, and also aiming at industry applications.

可以看出，中国科技巨头在AI视频领域的竞争已经进入了白热化的近身肉搏阶段，而对于创作者来说，工具越来越强，剩下的核心竞争力，或许真的只剩下...

It can be seen that the competition among Chinese tech giants in the AI video field has entered a white-hot stage of hand-to-hand combat, and for creators, as tools get stronger, the remaining core competitiveness might really just be...