Video generation models have a hard limit. Veo 3 gives you 8 seconds. Others give you 5. But stories don't stop at 8 seconds. By chaining generations together, you can create infinite, seamless 3D animations that look like a single continuous take.
The Loopback Technique
The concept is simple: take the end of one video and use it as the beginning of the next.
When generating a video, you get a start frame and an end frame. To extend the shot, you take the last frame of Video A and feed it into the "Start Frame" input of Video B. This forces the AI to start exactly where the previous video left off, creating a perfect visual stitch.
Repeat this process as many times as you want. 8 seconds becomes 16, then 24, then 32. Technically, you can go on forever.
Fixing the Audio
Here is the problem: AI models drift. Not just visually, but visually-auditorily. Even with the same prompt, the voice in Video B will sound slightly different from Video A. It breaks the immersion instantly.
To fix this, we use the Voice Clone node.
The full node workflow chaining multiple video generations.
The trick is to normalize the audio across all clips using a single target voice.
Take the audio track from your generated video and feed it into BOTH inputs of the Voice Clone node:
1. **Input A (Source)**: The video's audio track.
2. **Input B (Reference)**: The same video's audio track.
This forces the model to stabilize the voice, smoothing out the weird variations between generations. Finally, stitch all the video outputs together, and you have a single, coherent long-format animation.
Try the Workflow
We built a public workflow that implements this exact pipeline. Clone it and start experimenting.