Ever get frustrated trying to make AI videos with consistent characters? You generate a perfect image, but turning it into a video feels like a total gamble.
Well, I just stumbled upon an incredible post from Tianyu Xu that completely solves this. He’s figured out a super simple framework for prompting Veo 3’s new image-to-video feature, and I was blown away by its simplicity.
He calls it the CCAD framework. It’s a brilliant way to bring your static images to life with precision.
Here’s the breakdown Tianyu shared:
- 📸 [Camera] – Describe the camera motion (e.g., “the camera tilts down”).
- 🧍 [Character] – Mention the character in your image (e.g., “a husky in a podcast”).
- 🎬 [Action] – State what the character does (e.g., “takes out a notebook”).
- 💬 [Dialogue] – Add what the character says (e.g., “says: I found it in my notes”).
The magic here, as Tianyu points out, is that your source image does 80% of the work. This framework just gives you precise control over the animation and dialogue. No more random character changes!
I love his insight that you don’t even need all four parts every time. You can keep it simple depending on what you need.
This is such a practical hack for anyone diving into AI video. For the full breakdown and to see the videos he generated with this method, you HAVE to check out Tianyu Xu’s original post on LinkedIn!