Recherche avancée

Sur d’autres sites (15150)

• How come the quality of HEVC with alpha from my ffmpeg with videotoolbox is not ok

  27 juin 2023, par Jos

  I'm encoding ProRes 4444 with alpha channel video's to transparent video for web in both VP9 (webm) and HEVC (mp4). I'm using a great free tool by Rotato to encode to both formats at once.

  


  However, because I would like to have more control over e.g. the dimensions and the bitrate, I'm looking into ffmpeg. I can encode to both formats with the following params.

  


  # vp9
ffmpeg -i in.mov -vf scale=-1:720 -c:v libvpx-vp9 -auto-alt-ref 0 -pix_fmt yuva420p -acodec libvorbis -cpu-used 3 -deadline good -crf 40 out.webm 

# hevc
ffmpeg -i in.mov -vf scale=-1:720 -c:v hevc_videotoolbox -allow_sw 1 -alpha_quality 0.9 -tag:v hvc1 -q:v 65 out.mp4


  


  The problem : the videotoolbox encoder generates a weird white garbage around the black pencil animations and fade-in's from transparent to black pencil become fade-in's from white pencil to black pencil !?

  


  Here are screenshots of the two compared (top one is my ffmpeg output) :

  


  

  


  

  


  Anyone have good settings for ffmpeg I cano use to encode good quality HEVC with alpha ?
Hardware : MacBook Air M1 and Win11 + GTX 1060 (I should be able to use NVEnc, but don't know how)

  


• FFMPEG (WINDOWS) - Jerky Videos with vidstabdetect & vidstabtransform

  26 avril 2016, par Onish Mistry

  I require to stabilize multiple video clips and finally stitch all the clips, along with images into one final video. These "Scenes" consisting video clips as well as images also can have overlays like Texts and/or other Images.

  Basically the code I have in place as of now does everything for me just fine, where all the video clips are first converted into frame images. It then reads all the frames, puts on the overlays, adds a fade transition in-between "Scenes".

  Coming to the issue I am facing with stabilization, when I extract image frames out of the stabilized video clip and simply try to recreate video from those extracted image frames, it comes out with a weird jerk, almost like as if it is missing those stabilization calculations or something, not sure. It still looks a bit stabilized but with missing frames. I have checked duration and number of frames extracted, everything matches with the source, non-stabilized video.

  Below is the command used to stabilize the video, result of which is a perfectly stabilized video.

  ffmpeg -i 1.MOV -r 30 -vf vidstabdetect=result="transforms.trf" -f null NUL && ffmpeg -i 1.MOV -r 30 -vf vidstabtransform=smoothing=30:input="transforms.trf" -vcodec libx264 -b:v 2000k -f mp4 results.mp4

  Below is the command I use for video to image :

  ffmpeg -i results.mp4 -r 30 -qscale 1 -f image2 %d.jpg

  Below is the command I use for image to video :

  ffmpeg -i %d.jpg -r 30 -vcodec libx264 -b:v 2000k -f mp4 final.mp4

  Any help or suggestions are welcomed and appreciated.

  Thanks,

• fate : add chroma position scale test

  23 septembre 2015, par Christophe Gisquet

  fate : add chroma position scale test
  The sample position is made weird and non-nominal to force catching
  
  such issues as default values or specialized operations hiding
  
  issues in corner cases.
  Signed-off-by : Michael Niedermayer <michael@niedermayer.cc>
  

  • [DH] tests/fate/filter-video.mak
  • [DH] tests/ref/fate/filter-scalechroma