Movie Generation Benchmark Videos With Synchronized Audio

Encompassing non-cherry-picked generated videos from the Movie Gen benchmark, as detailed in the Movie Gen technical report. It is designed to enable fair comparison for future video generation research. The dataset includes high-quality 1080p HD videos with different aspect ratios and synchronized audio.

Use Cases

• Benchmark video generation models by comparing outputs to the non-cherry-picked generated videos.
• Evaluate the quality and synchronization of generated audio within the video files.
• Analyze the performance of generative AI models across different video aspect ratios.
• Use the high-quality 1080p HD videos as a reference set for multimedia generation tasks.

Strengths

• Includes high-quality 1080p HD video content.
• Videos feature synchronized audio, a key multimedia component.
• Dataset is specifically designed as a benchmark for fair comparison in generative AI research.

Limitations

• The specific number of videos, file formats, and total data size are unknown.
• No columnar or structured metadata is provided, limiting quantitative analysis.
• The dataset consists solely of synthetic, AI-generated content, which may not reflect real-world video distributions.

Provenance

Source

Meta AI for Media Research.

Collection Method

AI-generated videos produced by the Movie Gen foundation models for benchmark evaluation.

Freshness

Last updated on October 17, 2024.