You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Cropper which removes black bars from the videos before the merge (#50). It runs after the FPS changer and only when at least one video has bars. It samples about 66 frames spread over each video, so a black opening does not hide the bars, and it crops only where the bars are found. Videos without bars are copied unchanged, and when any video is cropped, all videos are converted so they can be merged. Rotated videos are compared in their displayed orientation. Can be skipped by --skip-cropper.
GPU video encoding for merging (#46). macOS uses VideoToolbox. On Linux and Windows, the first working GPU encoder (NVIDIA NVENC, Intel QSV, or AMD AMF) is used. If none works, the CPU encoder libx264 is used.
Fixed
Each input is now checked before the merge starts. If every input fits the output, they are copied as they are. If any input does not fit, all inputs are converted so their codecs match. MP3 output keeps MP3 audio (#102).
Chapter start times no longer drift earlier with each episode (#48). Chapter times are now stored in milliseconds, and each boundary is calculated from the running total of clip durations instead of whole seconds per clip.
Merging .mov inputs with a tmcd timecode stream no longer fails (#53). The timecode track is dropped with -write_tmcd 0, and if the merge still fails, each input is converted to MP4 and the merge is retried (#100).
Merging into MP4-family outputs no longer fails when an input has pcm_alaw audio (#17). The audio is re-encoded to AAC (-c:a aac) during the merge, because MP4 containers cannot store pcm_alaw.