Media Processing

FFmpeg58.3k★

FFmpeg is the invisible engine behind almost every video you've ever watched online. It converts, encodes, decodes, streams, and processes audio and video in every format imaginable. YouTube uses it. VLC uses it. Your favorite streaming service uses it. It's everywhere. If you need to process media programmatically — transcoding uploads, generating thumbnails, extracting audio, streaming to HLS — FFmpeg is not optional. There are no real open-source alternatives at this scope. HandBrake wraps FFmpeg for GUI transcoding. GStreamer is the pipeline-based alternative for real-time media processing. Commercially, cloud transcoding (AWS MediaConvert, Mux) abstracts FFmpeg away for a price. The power is completeness. Every codec, every container format, every filter. The command-line interface is a domain-specific language for media processing. The catch: the CLI is notoriously cryptic. Arguments interact in non-obvious ways, and the documentation assumes deep multimedia knowledge. The licensing is complex — depending on which codecs you compile in, you might be under LGPL or GPL, which matters if you're distributing software.

imgproxy10.5k★

The fastest open source image processing server you can self-host, and it's not close. In benchmarks, imgproxy outperforms Thumbor by 2x+ on both Intel and ARM, handling resize-on-the-fly with a security-first design. If you're building a content-heavy app and Cloudinary's per-transform pricing makes you wince, spin this up behind your CDN and forget about it. Thumbor offers more filters — blur, watermarks, face detection — but it's slower and harder to secure. Cloudinary is the polished SaaS option with advanced features, but vendor lock-in is real and costs scale fast. Imagor (also Go-based) is the closest alternative but has less community traction. Drop it in a Docker container, point your image URLs through it, and you get automatic WebP/AVIF conversion, smart cropping, and URL signing out of the box. The catch: imgproxy is a resizing server, not a full media pipeline. No face detection, no watermarks, no ML-powered smart features. If you need those, you're either stacking tools or going back to Thumbor.