Doubao Is Back with Another Epic Update

Even before the Spring Festival has officially begun, the battle among large AI models has already reached fever pitch. Competition across the entire ecosystem is intensifying at an astonishing pace, with new releases arriving almost back to back.

On the programming front, GPT-5.3-Codex and Claude Opus 4.6 have taken turns dominating the spotlight—one stepping aside just as the other makes its entrance—pushing code generation to ever higher standards.

The image-generation arena has been just as heated. Alibaba rolled out Qwen-Image-2.0, and almost immediately after, ByteDance followed up with Seedream 5.0 Preview, which not only improves image quality but also introduces online search integration, allowing visuals to be generated with the latest real-world information in mind.

Yet when it comes to truly standing out on a global scale—clearly ahead of peers by a wide margin—the spotlight over the past few days has belonged to the video-generation model that has swept across the internet:

Seedance 2.0.

The reason this discussion is happening now is simple. Until recently, Seedance 2.0 was still in closed beta. Free usage came with strict limits—very few points, just one or two videos per day—or required downloading an additional app. Either way, there was a noticeable barrier to entry.

What surprised everyone was how quickly ByteDance moved to scale it up.

Doubao has now officially announced that Seedance 2.0 is fully launched.

Open the Doubao app today and you’ll see a dedicated “Seedance 2.0” button at the bottom of the interface. Tap it, and you can generate videos directly from text prompts, or create videos using a real-person avatar.

The feature is available to everyone for free. When using the Seedance 2.0 model, users receive five free video generations per day—more than enough to explore and experiment.

That said, based on hands-on testing, there are currently fairly strict limitations around real-person video generation. You can’t simply upload someone else’s selfie or full-body photo and turn it into a video. The app verifies whether the subject is actually you.

You also can’t use arbitrary images as reference inputs, because Doubao does not yet support “reference image” generation. Given Doubao’s massive user base, it’s likely that opening all features at once would place heavy pressure on servers. Over time, these capabilities are expected to be gradually expanded.

As for just how high the model’s ceiling really is, many users have already seen the answer. Over the past few days, platforms like Bilibili, Xiaohongshu, and Douyin have been flooded with videos generated by Seedance 2.0.

The strongest takeaway is simple: everyone can be a director now.

Some creators have used Seedance 2.0 to produce fight scenes like “Jackie Chan vs. Jet Li,” complete with smooth, stable action, cinematic camera movement, and visuals that are nearly indistinguishable from live-action footage.

Others have gone even further. One creator placed Kobe Bryant into Arcane, seamlessly matching the show’s art style while delivering coherent action, shot composition, voice work, and visuals—without the painstaking editing that would traditionally be required.

This level of accessibility, technical leadership, and versatility has impressed even industry veterans.

Feng Ji, producer of Black Myth: Wukong, remarked, “The childhood era of AIGC is over.”

Renowned film director Jia Zhangke echoed the sentiment, saying, “I’m planning to use it to make a short film.”

For everyday users, however, the impact is even more tangible.

Many of us have imagined rewriting the ending of a favorite anime or TV series, creating AI-generated TV commercials for products, or even just watching our own cat battle an Ultraman for fun.

What once felt difficult or out of reach is now effortlessly achievable with Seedance 2.0.

The era of “if you want to see it, just make it yourself” has truly arrived.