v0.9.14

Changed

• Language detection now logs why a track stays und. Detection is fail-open, so failures (below-confidence guess, empty text/OCR extraction, silent mkvextract/ffmpeg errors) previously left a track und with no explanation. Each of those now logs a concise [LANG-DETECT]/[IMG-OCR] reason (with the confidence value, char/OCR counts, or subprocess stderr). Diagnostic only — no change to detection behavior.

Docker images

Pinned to this release:

docker pull ghcr.io/i-ial9000/shrinkerr:0.9.14         # portable (amd64 + arm64, CPU)
docker pull ghcr.io/i-ial9000/shrinkerr:0.9.14-nvenc   # x86_64 + NVIDIA GPU (NVENC)
docker pull ghcr.io/i-ial9000/shrinkerr:0.9.14-edge    # as above but ffmpeg master
docker pull ghcr.io/i-ial9000/shrinkerr:0.9.14-edge-nvenc

Floating tags (:latest, :edge, :nvenc, :edge-nvenc) track the most recent release of each lineage.