Stitching AI video clips together is a pain, right? You always get that awkward jump cut where the motion just doesn’t quite match.
Well, I just stumbled upon an absolutely brilliant post from Tianyu Xu that shares a super clever workaround for creating continuous, long-form AI video!
He figured out the secret sauce is using a model that supports start/end frames, like the new Kling v2.1.
Here’s the genius technique Tianyu Xu laid out:
- 🎬 Step 1: Create your first clip by generating a video from Frame A (your start) to Frame B (your end).
- ⛓️ Step 2: Take the very last frame of that clip (Frame B) and use it as the first frame for your next one.
- 🔄 Step 3: Generate the next clip from Frame B to Frame C, then C to D, and so on. Bam! You’re chaining them together seamlessly.
He even shared how you can create perfect loop videos this way: {Frame A → Frame B → Frame C → Frame A}. So cool!
Now, Tianyu Xu is quick to point out it’s not perfect yet. He notes that speed and movement can sometimes be inconsistent across the stitched clips.
But here’s his killer pro-tip:
we all have the attention span of a goldfish anyway! He says the real power is staying flexible. Leverage the frames, cut often, and don’t get hung up on creating one perfect, long shot.
This is a total game-changer. For all the details, go check out the original post from Tianyu Xu himself!