Workflow — AI Enhancement

What: Take existing footage and make it visually/sonically better using open-source AI. Strict license discipline: Apache-2 / MIT / BSD / GPL only. NC / research-only models are documented-and-dropped.

Skills used

media-upscale, media-interpolate, media-matte, media-depth, media-denoise-ai, media-demucs, ffmpeg-denoise-restore, ffmpeg-stabilize, ffmpeg-ivtc, ffmpeg-hdr-color, ffmpeg-lut-grade, ffmpeg-ocio-colorpro, ffmpeg-transcode, ffmpeg-hwaccel, ffmpeg-quality.

Pipeline

Step 1 — Probe source

ffmpeg-probe. Capture resolution, frame rate, codec artifacts, audio noise floor, color space (BT.601 / BT.709 / SDR), bit depth, chroma.

Step 2 — Undo legacy artifacts BEFORE AI