Ever feel like your AI videos are stuck in a silent movie?
Or you have a talking head that can’t actually DO anything? It’s a huge pain.
Well, I just stumbled upon an awesome post from Tianyu Xu that tackles this head-on. He’s been digging deep into Google’s new Veo 3 model, and I think what he’s found is a total game-changer.
According to Tianyu, most AI models can either show (visuals) or tell (lip-sync), but not both. Veo 3 does BOTH in the same clip! This is massive for anyone in education or storytelling.
He even shared his 5 golden rules for getting a higher success rate. I was blown away by the simplicity!
Here are Tianyu Xu’s tips for prompting Veo 3:
- 🎯 One Job Per Prompt: Don’t overload the AI by asking it to do too many things at once in a single prompt. Keep it simple and focused.
- ✍️ Use Plain English: No need for complex code or jargon. He says to describe the actions and effects clearly, just like you were explaining it to a person.
- 🖼️ Image First, Video Second: This was a brilliant insight! Tianyu advises putting most of your effort into describing the still image (the characters, the scene). If you nail that, the video part works much better.
- ✅ Embrace Imperfection: Don’t chase the perfect clip. He found that using good partial results is a much more effective workflow.
- 🛠️ Right Tool for the Job: Tianyu points out that Veo 3 has its limits. It’s not the best for overly wild visual effects, other models are better suited for that.
His takeaway is that Veo 3 is the ideal model for show and tell right now, and anyone can get the hang of it with these rules.
This is just my summary of his incredible work. For all the details and to see his examples, you HAVE to check out the original post by Tianyu Xu!