- Google launched Veo 3, an AI video synthesis model capable of generating synchronized audio tracks for video clips.
- The model can produce eight-second high-definition videos with voices, dialogue, and sound effects.
- A popular benchmark involves a humorous scenario of Will Smith eating spaghetti, first popularized in a viral AI-generated video from March 2023.
- Javi Lopez tested Veo 3 by recreating the spaghetti scene and noted an odd sound effect of crunching while Smith "eats."
- The glitch stems from the training data used by Veo 3, which included many examples of chewing sounds, affecting the accuracy of the generated audio.