Sora 2 vs. Veo 3: Who Comes Out on Top?
Sora 2 vs. Veo 3: Which AI creates better videos? See our head-to-head comparison of realism, audio sync, and prompt accuracy.
“AI Disruption” Publication 7800 Subscriptions 20% Discount Offer Link.
This time, OpenAI has pulled off something big again, unleashing Sora 2, which can directly generate 1080p videos up to 20 seconds long.
For example, some users have generated videos of Altman as a GPU delivery guy, and even scenes of stealing graphics cards from supermarkets.
Compared to the old version, Sora 2 excels in physical accuracy, realism, and controllability, and like Google’s Veo3, it also features audio-video synchronization capabilities.
We used the same prompts to pit both versions of Sora against each other.
Prompt: A black tech reviewer talking about a smartphone, while sitting at a desk in front of 2 displays.
The previous version of Sora didn’t truly “understand” the objects in the video, and the model was still prone to “hallucinations.”
The smartphone in the black guy’s right hand kept appearing and disappearing out of nowhere, his left hand would inadvertently gain or lose fingers, and the images on the displays also showed finger distortions.
