v0.7.14

Fixed

• NVENC conversions that crash mid-stream now auto-retry with software decode. NVDEC silently falls back to CPU decode for frames it can't handle on-GPU (sources with "unknown" colour metadata, certain WEB-DLs), and that mid-stream CUDA→CPU switch ("hwaccel changed") breaks the scale_cuda filter graph with Error reinitializing filters! (exit 218) — no input flag prevents it (v0.7.12's -noautoscale didn't). v0.7.14 detects that specific failure on the NVDEC-native CUDA path and retries the encode once with software decode + NVENC (software decoders handle every mid-stream change; the GPU still encodes). The fast NVDEC path stays the default for the 99% of files that work. Adds CI tests asserting the rebuilt command's structure (no scale_cuda on the software path; -noautoscale correctly positioned as an output option).

Docker images

Pinned to this release:

docker pull ghcr.io/i-ial9000/shrinkerr:0.7.14         # portable (amd64 + arm64, CPU)
docker pull ghcr.io/i-ial9000/shrinkerr:0.7.14-nvenc   # x86_64 + NVIDIA GPU (NVENC)
docker pull ghcr.io/i-ial9000/shrinkerr:0.7.14-edge    # as above but ffmpeg master
docker pull ghcr.io/i-ial9000/shrinkerr:0.7.14-edge-nvenc

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