Whoa, check this out. I just saw an incredible post from Tianyu Xu on how to break the length barrier for AI videos.
Ever get frustrated when your AI video generation caps out after a few seconds? Well, Tianyu figured out a super clever workaround using models that support start and end frames, like the new Kling v2.1.
I was blown away by how simple and effective his method is.
Here’s the breakdown of his technique:
⛓️ The Chaining Method
The secret is to create a series of shorter clips and stitch them together seamlessly.
- First Clip: You generate a video using your first image as the start frame (Frame A) and a second image as the end frame (Frame B).
🎬 = {Frame A → Frame B} - Chain It: Then, you take Frame B (the end of your first video) and use it as the start frame for your next video, which ends with Frame C.
🎬 = {Frame B → Frame C} - Repeat: You just keep doing this! {A → B → C → D…} You can theoretically make a video as long as you want.
💡 Bonus Tip from Tianyu: Loop Videos!
He also points out you can create perfect loops by making your very last end frame the same as your first start frame. How cool is that?
🎬 = {Frame A → B → C → A}
⚙️ A Quick Reality Check
Tianyu is quick to note that it’s not perfect. The speed and movements can be a bit inconsistent between the clips you chain together. It’s a workaround, not a perfect feature.
But here’s his killer insight…
He reminds us that the average human attention span is just 8 seconds (a goldfish has 9! 🐠).
So, for social media, super long, continuous shots are rarely the answer anyway. His advice is to stay flexible: leverage individual frames, cut often, and edit freely.
This is a game-changing way to think about AI video creation. For the full scoop, go check out the original post from Tianyu Xu himself!