finds.dev← search

// the find

CraigWatt/vfo

★ 5 · C · GPL-3.0 · updated Apr 2026

Autonomous media optimization engine for video libraries: mezzanine→source→profile workflows with quality checks and observability.

vfo is a C CLI tool that wraps FFmpeg with a rule-based profile system — you define what a target output should look like, and it figures out which FFmpeg command to run for each file based on codec, resolution, color space, and other properties. It's aimed at people managing large personal media libraries who are tired of writing one-off encode scripts for every edge case.

The profile/scenario model is the right abstraction for heterogeneous libraries — instead of a single preset, you get conditional dispatch that handles 'file is already correct, just remux' vs 'downscale this 4K file' vs 'this has no valid color space' as separate cases. The VMAF/PSNR/SSIM quality gating is real and configurable, not just a checkbox — you can set minimum thresholds and fail the encode if they're not met. The observability layer (status, status-json, visualize with Mermaid/HTML output) means you can actually see what the pipeline decided and why, which matters a lot when debugging a mismatch between what you expected and what got encoded. The wizard with preset packs for Roku, Fire TV, Chromecast, Apple TV, and Dolby Vision Fire Stick is a meaningful time-saver — those device compatibility matrices are annoying to maintain by hand.

Five stars and zero forks means this is essentially a personal tool that got published, and the documentation is honest about it — macOS is the only fully supported path, Linux and Windows are explicitly marked work in progress. The config system is a flat key-value conf file with a naming convention that grows linearly as profiles multiply; there's no validation schema, so a typo in QUEEN_CRITERIA_CODEC_NAME silently does nothing. The AGENTS.md / Codex autonomous loop CI wiring is AI-assisted development scaffolding baked into the public repo, which is an unusual thing to ship and suggests the project is as much an experiment in AI-augmented workflows as it is a finished media tool. There's no daemon or watch mode — this is a run-it-manually-or-from-cron tool, which limits its usefulness if you're ingesting media continuously.

View on GitHub → Homepage ↗

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →