I love AI-generated content, so I was excited to try Google's Gemini Veo 3 to make videos. Here's what I learned from using this tool.
When I first tried Veo 3, I liked that it creates both videos and matching audio. My first simple prompts gave interesting results, but they weren't as good as I hoped.
I soon learned that writing good prompts takes practice. After trying different approaches, I found that including these elements made my videos much better:
My biggest improvement came when I focused on facial expressions. Using the word "portrait" and asking for specific emotions made the AI create more realistic human expressions. This worked great for videos of people reacting to things.
For better audio, I found three important things to include:
One of my best videos showed two children with different emotional responses. By including their thoughts and background sounds, the video felt much more real.
A slow-motion video showing two Asian children with different reactions to conflict. One child stands tall with confident posture, while the other appears anxious and uncertain. The scene transitions between them showing their contrasting emotional responses.
Camera starts with a cinematic split-screen showing both Asian children simultaneously, then zooms in on each child's face to capture their emotional expressions. Soft focus on their eyes reveals their inner thoughts.
The first child (representing "knife edge") demonstrates resilience with decisive body language and calm facial expressions when confronted. They bow slightly but maintain eye contact, showing respect while preserving boundaries. The second child (representing "sponge") shows absorption of external stress, visibly overthinking and second-guessing themselves, repeatedly bowing and looking down.
Style:
- Documentary-like with warm lighting. Background ambient sounds of traditional Asian string instruments that shifts in tone between the two characters.
- Subtle dialogue: "没关系，再见" (Méiguānxì, zàijiàn - "It's fine, goodbye") from the confident child versus
- Whispered internal monologue "我做错了吗？我应该再道歉一次吗？" (Wǒ zuò cuò le ma? Wǒ yīnggāi zài dàoqiàn yīcì ma? - "Did I do something wrong? Should I apologize again?") from the anxious child.
Example video generated by Gemini Veo 3
Example video generated by Gemini Veo 3
A slow-motion video showing two children with different reactions to conflict. One child stands tall with confident posture, while the other appears anxious and uncertain. The scene transitions between them showing their contrasting emotional responses.
Camera starts with a cinematic split-screen showing both children simultaneously, then zooms in on each child's face to capture their emotional expressions. Soft focus on their eyes reveals their inner thoughts.
The first child (representing "knife edge") demonstrates resilience with decisive body language and calm facial expressions when confronted. The second child (representing "sponge") shows absorption of external stress, visibly overthinking and second-guessing themselves.
Style: Documentary-like with warm lighting. Background ambient sounds of soft piano music that shifts in tone between the two characters. Include subtle dialogue: "It's fine, goodbye" from the confident child versus whispered internal monologue "Did I do something wrong? Should I apologize again?" from the anxious child.