Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V

Key points

• Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation.1
• We propose a dual-level evaluation framework combining 12 objective metrics and five subjective metrics.1
• Extensive evaluation of 15 recent proprietary and open-weight MLLMs from nine families reveals that even leading models face challenges in reliably generating high-quality complete videos aligned with textual and visual requirements, with performance generally declining as construction, dynamic animation, and state-dependency complexity increases.1
• We anticipate that Code2Video-Compass will advance reliable, coherent, and controllable code-based video generation and inspire broader progress in programmatic visual creation.1

Sources (1)

• [1]Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 03:51 PM

  We introduce Code2Video-Compass, a general-purpose benchmark for code-based video generation across text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V) settings.

  Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation.

Extractive summary: sentences quoted from the sources.