MAT_S01: Scene-Level Video Materials for AI-Generated Style Analysis

Jinyu Kim's doctoral dissertation project provides scene-level video materials for a comparative analysis of original and AI-generated videos. The dataset includes RAW and PRE (preprocessed) videos, prompt files, generation settings, and metadata for the 1926 Soviet montage film 'Mat' (Mother) by Vsevolod Pudovkin. The PRE files are standardized at 720p, 24 fps, and 8 seconds, with audio excluded from analysis.

Use Cases

• Compare original and AI-generated video styles based on scene-level materials.
• Analyze the reconfiguration of film styles based on the provided Soviet montage case study.
• Study generative AI video model outputs based on the included RAW generated videos and prompt files.
• Benchmark video preprocessing pipelines based on the standardized 720p, 24 fps PRE files.

Strengths

• Includes both RAW generated videos and standardized PRE videos at 720p, 24 fps, 8 seconds.
• Contains prompt files and generation settings, which are crucial for reproducibility in AI video analysis.
• Scene-level organization and metadata for a specific film (Mat, 1926) provide a focused case study.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale statistical analysis.
• Column-level documentation is absent; field semantics must be inferred after download.
• Audio is excluded from the analysis, limiting use cases requiring audiovisual study.

Provenance

Source

figshare, author Jinyu Kim.

Collection Method

Created as part of a doctoral dissertation project on film style reconfiguration via generative AI.

Time Range

Based on the 1926 film 'Mat'.

Freshness

Last updated 2026-06-04 21:01:50; freshness should be verified.