Seedance Strategy

TBPN，科技每日秀 Substack 专栏。关注

Happy Monday.

We’re live on YouTube and 𝕏.

The current thing in tech and business is Amazon blocks Meta’s Muse.

Today’s Lineup

Rainmaker Co-Founder & CEO Augustus Doricko at 11:45 AM
Khosla Ventures Founder Vinod Khosla at 12:00 PM
Box Co-Founder & CEO Aaron Levie at 12:20 PM
TomorrowX Summit Co-Host Ron Biscardi at 12:50 PM
Raindrop Co-Founder & CTO Ben Hylak at 1:00 PM

Run of Show

Seedance Strategy

The latest turn of AI video models is really really good. Seedance 2.5 can easily replace a whole person (not just a face swap) and match all the subtle details of the footage so things integrate really nicely. It really seemed like people fell for Dylan Patel’s swapped Drake Cheetah Print video over the weekend.

It’s very easy to make these now, on Sunday I had Codex finding famous or funny videos to use as base clips, and then managing the whole process of passing Jordi and I into the model for character replacement. I never had to leave a text box, and I didn’t even have to upload reference photos because I could just say go get them from the internet or photoshoots I have stored in my Google Drive. The results are really good.

So, expect lots of memes, the usual “it’s over” cycle for traditional workflows, and then obviously a whole lot of blended work that uses the right tool for the job. Some productions will use this where it helps in urgent scenarios. The famous example is Henry Cavill’s Mission Impossible mustache having to be removed in post-production with some very tedious and noticeable CGI because of reshoot scheduling conflicts.

So, fun model from Seedance 2.5, but I had some questions about the relative strategy and market structure here. We know that at a high level, intellectual property rules are… somewhat looser in China, but these models are insanely compute intensive, and America is way ahead on compute buildout. So what’s going on?

The biggest piece of the answer is applications and alignment with core business. ByteDance owns an enormous video business across TikTok/Douyin, CapCut, plus ads, e-commerce, recommendations, etc. Immediate benefit to the core business, without needing to pull together a new demand surface for video. This sort of exists in America with Meta and Google/YouTube, but AI “slop” is deeply unpopular and both companies are chasing coding / agents / RSI.

IP infringement isn’t dead to rights at this moment, but there is very strong evidence that Seedance learned copyrighted Hollywood characters to a remarkable degree. At the very least, lots of clips from Western cinematic IP make their way onto TikTok, if not whole movies moving through ByteDance servers, even if they are taken down at some point.

Seedance is not open-source or open-weight, in somewhat of a narrative violation to the default Chinese AI strategy. There are a few reasons for this. First, no one wants to open source when they are in the lead, we’ve seen this time and time again. Second, with coding models, open-weights are a play to build a Chinese technology stack, but for video, ByteDance already owns the distribution. Video plays on Chinese video apps that run on phones made in China. When you’re a challenger that needs an ecosystem, opening up makes sense, if you are an incumbent with distribution, you have to close down.

Now there are still some interesting knock on effects, mainly around speculative future value of video models in training humanoid robotics, but that’s further off. At least for the next few months, I just expect a bunch of fun videos to burn through the internet until it gets old.