Jasna Video Restore
AI neural video restoration workbench: upload → auto scan → TensorRT inference → outputs, processed locally on an AIPod compute pod.
Jasna is an open-source neural video restoration engine (RF-DETR detection + BasicVSR++ restoration, AGPL-3.0; this repo jasna-aipod is the official porting layer). This app ports it fully onto a LazyCat AIPod (NVIDIA Jetson AGX Orin): videos are processed entirely on the local pod, never uploaded to any cloud. Porting highlights: - Full workbench: chunked upload → auto scan → task queue (segment/full render) → output library with version chains; - Compare preview: side-by-side source vs restored playback; - Save back: outputs can be saved to LazyCat Drive (free directory choice) or downloaded locally; - First-run demo: one-click tour with a built-in 10s test clip (~30s end to end); - Jetson optimized: NVDEC/NVENC hardware codecs, TensorRT fp16, prebuilt engines for zero-compile startup, encode watchdog and resumable tasks. Open source: github.com/vito01111/jasna-aipod (with upstream credit to Kruk2/jasna).
