“A muscular barbarian breaking a CRT tv set with a weapon, cinematic, 8K, studio lighting”
“A scared girl in a Victorian outfit working by a forest, dolly shot”
“Low angle static shot: A teddy bear sitting on a picnic blanket in a park, consuming a slice of pizza. The teddy bear is brown and fluffy, with a pink bowtie, and the pizza slice is gooey with cheese and pepperoni. The solar is setting, casting a golden glow over the scene”
“Aerial shot of a small American city getting deluged with liquid cheese after a large cheese rainstorm the place liquid cheese rained down and dripped all around the buildings”
Additionally, we added a brand new one: “A younger girl doing a fancy flooring gymnastics routine on the Olympics, that includes working and flips.”
Weighing the outcomes
Total, the outcomes proven above appear pretty similar to Gen-3 Alpha and Minimax video-01, and that is notable as a result of HunyuanVideo may be downloaded at no cost, fine-tuned, and run regionally in an uncensored approach (given the suitable {hardware}).
There are some flaws, in fact. The vaudeville robots usually are not animals, the cat is consuming from a bizarre clear beer can, and the person consuming spaghetti is clearly not Will Smith. There seems to be some movie star censorship within the metadata/labeling of the coaching knowledge, which differs from Kling and Minimax’s AI video choices. And sure, the gymnast has some anatomical points.
