نسخة أولية وصول مفتوح
Dreaming the Sound of Contact: Leveraging Video and Audio Generation for Zero-Shot Force-Aware Manipulation and Data Generation
Video generation models have advanced rapidly and can now synthesize plausible videos of robot manipulation from image and text prompts. Recent work extracts robot actions directly from such generated videos, but the result is purely kinematic and lacks force information, causing failures in contact-rich tasks where ap …