Recherche avancée

Sur d’autres sites (7884)

• Anomalie #4043 : table debug-nav en partie masquée

  15 septembre 2018, par b b

  Fichier au format patch pour y voir plus clair.

• Color spot in FFmpeg output

  2 juin 2022, par Fabian Garcia

  I work in a company that serves multimedia content to different clients and one of these requests that the videos contain the following specifications :

  


  

  • Codec : H264 High@4.0

  


  • Mode : CBR 1 pass

  


  • Window Buffer : 5 secs

  


  • Frame Rate : 29.97/23.97

  


  • Key Frame Interval : 2 secs

  


  • Bitrate : 4000 Kbps

  


  • Size : 1920x1080

  


  • Window Buffer : 5 secs

  


  • GOP : closed

  


  • GOP size : fixed (no scene change detection)

  


  • Window Buffer : 5 secs

  


  • B Frames : As needed

  


  


  For the above, i build following script :

  


  ffmpeg -hide_banner -nostats -loglevel error -y
-init_hw_device cuda=cuda -hwaccel cuda -hwaccel_device cuda
-i 'video_source.mxf' -filter_hw_device cuda
-filter_complex '[0:v]format=nv12,hwupload,yadif_cuda,scale_cuda=1920:1080[v0]'
-map '[v0]' -c:v h264_nvenc -rc cbr -b:v 4M -bufsize 8M -x264-params "keyint=60:min-keyint=48:no-scenecut" -r '30000/1001' -profile:v high -level:v 4
-map 0:1 -c:a libfdk_aac -b:a 96k -ar 44100 -ac 2 -sample_fmt s16
'output.mp4'


  


  The inconvenience is that the generated output presents some color spots in the scene changes (it's a fade) as can be seen in the attached image.

  


  I could notice that in the output video, the time does not match the input precisely, the output is a few milliseconds ahead, I think that the color spot corresponds to the image that should be shown at that moment in the original video.

  


  Can you see something wrong in script ? Or maybe some concept that i don't know is causing a conflict with the output (This is my first time using ffmpeg).

  


  Thank for your help !

  


  

  


• ffmpeg - How to resize frame collage (grid of images) to grid of different dimensions, while respecting frame order ?

  12 mars 2023, par JamesK

  I have an input of a "frame collage" (that is, frames extracted from a video, arranged in a certain grid). I want to rearrange the frames from a collage into a different grid shape. I can calculate the input coordinates (or bounds, e.g. x.min -> x.max, y.min -> y.max) and output coordinates for each frame. Basically, I want to resize a grid, while respecting the positioning of each element of the original grid.

  


  As an example, given a frame collage that is 9x2 (9 columns wide, 2 rows tall), with image resolution of 1080x240 (width x height), resize the frame collage to be 6x3 (6 columns wide, 3 columns tall), with an image resolution of 720x480 (width x height), while respecting the order of the frames in the input image (left to right, top to bottom).

  


  I generated the frame mappings (where the original frame was located, and where the new frame should be placed) for each frame, a table of that data is available here : https://pastebin.com/raw/a5UhKhek (posted on pastebin because discord would murder the formatting).

  


  Does anyone have any ideas on how I might be able to pull this off ?

  


  I've investigated the tile and untile filter, as well as looked into the select filter, although the syntax for the lattermost is a touch confusing to me.