On Tuesday, Google Veo 3 launchedA new video synthesis model of artificial intelligence can do something that the main video generator has not been able to artificial intelligence before: Create a simultaneous audio path. While from 2022 to 2024, we saw early steps in generating video from artificial intelligence, every clip was silent and usually very short in the period. You can now hear the sounds, dialogue of dialogue and sound effects in eight -second -definitive videos.
Soon after the new launch, people started asking the most obvious standard question: How good is Veo 3 in the mixture of an Oscar -winning actor Will Smith in eating spaghetti?
First, a short extract. The spaghetti standard in AI video tracks its origins to March 2023, when we first covered an early example on the video of artificial intelligence using an open source video creation form called Modelscope. The example of spaghetti later became known enough that Smith mocked this nearly a year later in February 2024.
Here’s the original viral video:
One thing that forgets people is that at the time, the example of SMTH Running It has already achieved superior results (although it was not yet available to the public). But the outcome of the models was funny and strange enough to stick to the memories of people as an early poor for synthesis of video, beneficial to future comparisons as artificial intelligence models are offered.
The developer of the Ai Javi Lopez application came for the first time to save the curious spaghetti fans earlier this week with the Veo 3, where they tested Smith and Publish the results On X. But as you will notice below when watching it, the soundtrack enjoys curiosity: it seems that Smith faux wanders in spaghetti.
On X, Javi Lopez has played “Will Smith Eating Spaghetti” in Google Veo 3 Ai Generator and got this result.
It is a defect in the experimental ability of VEO3 to the application of audio effects on the video, most likely because the training data used to create an Amnesty International Models of Google are characterized by many examples of chewing mouths with sound effects. The Model I intelligence models are the prediction machines that match patterns, and sufficient examples of different types of media must appear to generate convincing new outputs. If the concept is excessively represented or represented in training data, you will see unusual obstetric results, such as Jabberwockies.
