video-comparer_skill

Overview

This skill compares two videos and generates a self-contained, interactive HTML report to analyze compression results and visible quality differences. It extracts video metadata, computes PSNR and SSIM, and produces frame-by-frame visual comparisons with slider, side-by-side, and grid viewing modes. The report is offline-ready and includes zoom controls and aggregated quality summaries.

How this skill works

The tool validates inputs and probes each video with FFprobe, extracts frames at a configurable interval using FFmpeg, and scales frames to a consistent height for fair comparison. It computes pixel-level (PSNR) and perceptual (SSIM) metrics per frame and aggregates them across the clip. Frames are embedded as base64 images into a single HTML file and temporary files are cleaned automatically. Security checks prevent path traversal, command injection, and enforce file-size and timeout limits.

When to use it

• When you need to compare an original video against a compressed or transcoded version
• To evaluate codec choices, bitrate reductions, or parameter tuning impact on visual quality
• When producing before/after reports for quality assurance or delivery sign-off
• For automated batch comparisons across multiple assets or versions
• When you want a portable, shareable interactive report without running a server

Best practices

• Install and verify FFmpeg/FFprobe on the host before running comparisons
• Use the same resolution and color space for both videos when possible to avoid misleading metrics
• Increase frame extraction interval for long videos to reduce processing time, or lower it for critical segments
• Keep input files under the configured size limit or adjust the constant if you control the environment
• Run batch jobs in a controlled environment with sufficient CPU and a timeout buffer to avoid truncation

Example use cases

• Compare a master file and its CDN-transcoded output to confirm acceptable quality loss
• Benchmark two encoders or bitrate ladders by generating side-by-side metric summaries
• Create a client-facing report showing visual differences after applying a new compression profile
• Automate nightly comparisons of recently encoded assets to detect regressions
• Generate compact QA artifacts for bug reports that include both metrics and visual evidence