Seedance 2.0 global release: Everything we know so far
ByteDance hits a legal wall, Runway pivots its entire strategy, and Pika wants to clone you.
Good day! Grab a coffee and settle in. It’s been a wild week in the world of pixels and frames. Here is the latest on what’s happening in the AI world.
1️⃣ Seedance 2.0: The giant that can’t leave China (yet)
2️⃣ Runway pivots to provider of AI Models
3️⃣ Pika AI “Selves”: your digital doppelgänger
4️⃣ Quick news you may have missed
5️⃣ AI Video of the Week
Let’s go.
Seedance 2.0: The giant that can’t leave China (yet)
Everybody was waiting for it. ByteDance’s Seedance 2.0 was coming, but if you’re outside of China, start tempering your hopes. While the model is officially live on Jianying (the Chinese version of CapCut), the planned global rollout for February 24 has been postponed indefinitely.
Why? A massive legal wall. Hollywood heavyweights like Disney, Netflix, and the MPA issued legal notices after AI clips of Stranger Things and Star Trek went viral. ByteDance is now back at the drawing board, promising stricter filters for IP and deepfakes before we get a turn.
Why it matters: It proves that AI video has basically solved “physics,” but it also shows that the era of “train now, ask for forgiveness later” is officially over. We’re entering the age of “Clean AI” vs. “Cool AI,” where your favorite tools might be “nerfed” before they even launch.
Runway pivots to provider of AI Models
In a massive strategy shift, Runway is moving away from being a builder of a model to becoming a model provider like Freepik, Krea and Higgsfield. Their platform now hosts a lineup including Kling 3.0, Sora 2 Pro, and the new WAN 2.2 Animate.
Why it matters: They are now focused on building the best environment for creators, instead of fighting with Google and ByteDance for the best model.
Pika AI “Selves”: your digital doppelgänger
Pika just introduced Selves, a feature that lets you create a persistent AI version of yourself. This isn’t just a face-swap; it’s a multimodal agent that can talk, post, and “remember” things across platforms like Slack and Discord. It’s designed to learn your personality over time so it can represent you in video and text without you needing to be there (that line is scary and amazing at the same time).
Why it matters: We’re moving from “making cool videos” to “building digital identities.” It’s the first step toward having a digital assistant that actually looks and sounds like you, rather than a generic robot. Or you know, replace you :)
Quick news you may have missed:
Google Lyria 3: Released this week, it now allows you to generate 30-second soundtracks directly inside the Gemini app.
Claude Code + Figma: A new “Claude Code to Figma” integration is here, letting developers turn code into high-fidelity designs (and vice-versa) via the Model Context Protocol (MCP).
Midjourney v8: David Holz confirmed they are in the “polishing phase,” with a release expected any day now. They’re currently running “rating parties” to fix how the model handles text and typography.
AI Video of the Week
In the tsunami of Seedance 2.0 videos, one stood above the rest. The Dor Brothers went for the biggest blockbuster of all, spending billions… of credits in this video.
There you go. Have fun, touch grass, and create awesome stuff.
Thanks Erik, this is such a great post. The ByteDance commentary in particular is something that I think we all need to keep a close eye on. I honestly wonder, though, if major organisations are going to be able to keep up with the litigations, or if it's just going to lead to a fundamental shift in how we do copyright and IP going forward.