AI video generators are evolving rapidly, but I have always found them more interesting than convenient. In most cases, the clip collapses under scrutiny. The characters move awkwardly, the textures don't hold up, and the eerie valley takes over. They are fun experiments, but rarely persuasive.
Google's Gemini Veo 3 has rethinked that stance. Unlike other tools, it is not limited to text prompts. You can also extend still images to movement and add soundtracks. To test the VEO 3, I purposely gave it a diverse set of prompts, ranging from whimsical to creepy. Some results surprised me with their Poland and vibes, while others reminded me why AI videos still have limitations.
If you want to try Veo 3 yourself with Gemini, the process is simple. Log in to Gemini and[ビデオ]Click In the prompt box. Here are the three prompts I used:
1. Fluffy duck family
Please take a look
For this prompt, I wanted to see what Veo 3 is based on realism: how well it handles animals. Ducks are a solid test subject due to their extremely complex features and their layered feathers, reflective water and distortion.
I used the following prompt: “Build a fluffy duck family walking along the riverbank.” I was surprised at the Veo 3 output. The duck was immediately recognizable and had persuasive fluffy feathers.
But what really stood out was the environment. The mud and grass along the edge of the pond had a surreal texture with the subtle depth and variation you would expect from the actual footage. Light light refracted naturally across the water, even getting caught up in the treetop. This was a really nice touch.
The duck outlines sometimes softened, and at some point it didn't look like a duck. But despite those quirks, the scenes overall were far more convincing, refined, cohesive and far more convincing than I thought.
2. Fun fair and scary clown
Please take a look
For the second test, I wanted to push the Veo 3 creepy and surreal. The idea was simple: “Create a fun, fair and scary clown that tides with hair and people's cabbage juice.” It's a kind of prompt that can easily fall into a cartoon-like parody if AI struggles with tone.
Instead, the Veo 3 nailed it. The costume and makeup looked surprisingly realistic, down to the shaky look that wobbled between the comic and the ominous. And the vintage style Combs made the scene feel even more quirky. That touch of peculiarity gave the video a strange authenticity.
The only stumbling came when the clown squealed cabbage juice as the liquid didn't squeal from the opening of the bottle. Oddly, I think Glitch works in that favor and leaning against the scene's chaotic, circus-like energy.
The results were surreal and surreal. It was exactly the balance that Veo 3 wanted to attack.
3. Aliens driving on a spaceship
Please take a look
After recently watching “Alien Earth,” I had an extraterrestrial Earth in my brain. So I wanted to see how VEO 3 handles something more imaginative.
I used this prompt: “Create a hyper-realistic video of an alien stalker passing through a dark corridor in a spaceship towards the camera. The camera must be from our perspective.” The test here is not only about the design of creatures, but also about how persuasive the interior settings are.
The result really put me on the floor. The video played like a movie trailer, with the aliens sprinting straight to the camera before stopping several inches away. Its tail swayed with real weight, selling fantasies of momentum, and the ship's lighting created an atmospheric metal background that felt real science fiction.
What impressed me the most was the design itself. The creature looked like a mashup of familiar icons – Part Xenomorph, Part Doctor Foods Ord, and its humanoid-like torso properly measured the predator's touch.
4. Rock band music video
Please take a look
About one of the tougher tests, I asked Veo 3 “Give me a music video of a pop rock band performing on stage in front of a crowd.” The idea was to see how well you could capture performance energy, not just realism.
At first it seemed promising. The fireworks and staging were very convincing, and the overall framing had a proper concert vibe. But things were quickly unraveled.
The lead singer's face literally changed the headpan in the center before returning again. The hands in the crowd looked blurry and yankee, lacking the texture and details that made the previous clip work very well.
Even the audio that the Veo 3 automatically stacked was missing. The backing track was fun, but the lyrics (and the vocals for that) did not match the prompts provided. And, even worse, the strobe light remained completely static.
5. Mother and baby
Please take a look
For the prompt, I wanted to make something more intimate “Make a mother playing with her newborn baby at home and sit on the couch.” Compared to aliens and clowns, this should be easy. A test of realism, warmth and everyday details.
At first glance, the still image looked spot-on. Soft amber light filtered into the room, the mother's hair and clothing had an incredible texture, and the living room itself looked compelling. However, as soon as the video played, a crack appeared. Both mother and baby characteristics were blurred and blurred whenever they moved, completely destroying the illusion.
The overlayed sound was no use either. The piano track played over the clip, but the mother's dialogue bleed. When she said Ah, my sweet little thing, It should have been the emotional anchor of the scene. Instead, the competing music felt confusing and distracting.
What was a soft, light-electoral moment turned into something creepy. It is evidence that Veo3 may still struggle when subtle human emotions are focused.
to follow Tom's Guide on Google News To get the latest news, how-tos and feed reviews. surely[フォロー]Please click the button.
More from Tom's Guide
I'm back to my mobile phone
