media-processing_skill

Overview

This skill processes video, audio, and image files using FFmpeg and ImageMagick to handle conversion, encoding, filtering, streaming, and batch image manipulation. It supports hardware acceleration, 100+ formats, complex filtergraphs, and common streaming workflows (HLS/DASH). It is optimized for tasks like thumbnails, format conversion, codec-specific encoding, and image composition.

How this skill works

The skill constructs and runs FFmpeg commands for video/audio tasks: encoding, transcoding, frame extraction, streaming manifest generation, and hardware-accelerated encoding via NVENC/QSV. For images it uses ImageMagick (magick/mogrify) to convert formats, resize/crop, apply effects, watermark, and batch-process large sets. It also inspects media with ffprobe/identify to drive conditional pipelines.

When to use it

• Convert or transcode videos and audio between formats and codecs (H.264, H.265, VP9, AV1).
• Generate thumbnails, previews, or extract audio tracks from video files.
• Create HLS/DASH manifests and prepare assets for adaptive streaming.
• Batch resize, optimize, or apply effects to large image collections.
• Apply overlays, watermarks, or composite images/videos programmatically.
• Optimize media for web delivery with CRF, presets, and faststart flags.

Best practices

• Prefer -c copy when no re-encoding is needed to save time and preserve quality.
• Use CRF for consistent quality control; choose presets to balance speed vs. compression.
• Test settings on small samples before batch processing large datasets.
• Enable hardware acceleration (NVENC/QSV) for large encodes to reduce CPU load.
• Strip metadata and use progressive JPEG/optimized WebP for web delivery.
• Parallelize independent jobs (GNU Parallel) but cap threads to avoid memory exhaustion.

Example use cases

• Transcode a set of .mov files to H.264 MP4 with consistent CRF and audio bitrate for web publishing.
• Generate 4 responsive image sizes from a master image and place them in a CDN-ready folder.
• Produce HLS renditions with multiple bitrates and an m3u8 playlist for adaptive streaming.
• Extract a 30-second clip and a thumbnail at 5s for a video preview card.
• Batch-convert PNG assets to optimized JPEGs and strip metadata for size reduction.