Google’s Will Smith double is better at eating AI spaghetti … but it’s crunchy?
1 min read
Summary
Google has launched Veo 3, the first AI video synthesis that creates a synchronized audio track, meaning it includes voices, dialogue and sound effects in eight-second HD videos.
People are inquisitive as to how good Veo 3 is at generating footage of Oscar-winning actor Will Smith eating spaghetti, a famous benchmark for AI video generation.
At the end of 2023, Smith parodyed an early AI-generated video of him eating spaghetti, which was created with an open-source video synthesis model called ModelScope.
-AI app developer Javi Lopez has reportedly used Veo 3 to attempt the Will Smith test, with the results posted on X.
The video generates audio mouth sounds as the AI mouth movements synch with the audio, “crunching” sounds as the AI mouth moves, which itself is generated as the AI “looks” for synched audio.
This all highlights that AI models need to be trained on sufficient representative data to avoid unusual generation results.
