Recherche avancée

Sur d’autres sites (10457)

• How to scale and mux audio ?

  20 décembre 2019, par siods333333

  First problem is with audio rescaling. I’m trying to redo doc/examples/transcode_aac.c so that it also resamples from 41100 to 48000, it contained a warning that it can’t do it.

  Using doc/examples/resampling_audio.c as a reference, I saw that before doing swr_convert, I need to find the number of audio samples at the output with the code like this :

      int dst_nb_samples = av_rescale_rnd( input_frame->nb_samples + swr_get_delay(resampler_context, 41100),
  
                                           48000, 41100, AV_ROUND_UP);

  Problem is, when I just set int dst_nb_samples = input_frame->nb_samples (which is 1024), it encodes and plays normally, but when I do that av_rescale_rnd thing (which results in 1196), audio is slowed down and distorted, like there are skips in the audio.

  Second problem is with trying to mux webm with opus audio.

  When I set AVStream->time_base to 1/48000, and increase AVFrame->pts by 960, the resulted file is played in the player as a file that is much bigger. 17 seconds audio shows as 16m11s audio, but it plays normally.

  When I increase pts by 20, it displays normally, but has a lot of [libopus @ 00ffa660] Queue input is backward in time messages during the encoding. Same for pts 30, still has those messages.

  Should I try time_scale 1/1000 ? webm always have timecodes in milliseconds, and opus have packet size of 20ms (960 samples at 48000 Hz).

  Search for pts += 20;

  Here is the whole file, all modification I did are marked with //MINE : http://www.mediafire.com/file/jlgo7x4hiz7bw64/transcode_aac.c

  Here is the file I tested it on http://www.mediafire.com/file/zdy0zarlqw3qn6s/480P_600K_71149981_soundonly.mkv

• ffmpeg scale doesn't work if destination res is larger than source

  21 décembre 2017, par Vibok

  I just tried ffmpeg filter scale=320x180:force_original_aspect_ratio=disable to resize a 270x480 video into a 320x180 video, and it still keeps aspect ratio. force_original_aspect_ratio=disable got ignored.

  I guess the problem is that the destination width is bigger than the source width, while the destination height is smaller. Because it worked for other videos, even without force_original_aspect_ratio=disable.

  The resulted file is also weird. It says it’s 320x180, while its weight is obliviously smaller than that.

  Here are the video files, original and resized. https://drive.google.com/file/d/1UNXlfwpzoizhx7WOjqn44mlcbQgacOHS/view?usp=sharing

  Here is my command :

  ffmpeg -i 480P_600K_107047752.mp4 -force_key_frames 00:00:03.000 -filter_complex [0:v]scale=320x180:force_original_aspect_ratio=disable,fps=30[vid];[vid]

• Workflow for creating animated hand-drawn videos - encoding difficulties

  8 décembre 2017, par Mircode

  I want to create YouTube videos, kind of in the style of a white-board animation.

  Tldr question : How can I encode into a lossless rgb video format with ffmpeg, including alpha channel ?

  More detailed :
  My current workflow looks like this :

  I draw the slides in Inkscape, I group all paths that are to be drawn in one go (one scene so to say) and store the slide as svg. Then I run a custom python script over that. It animates the slides as described here https://codepen.io/MyXoToD/post/howto-self-drawing-svg-animation. Each frame is exported as svg, converted to png and fed to ffmpeg for making a video from it.

  For every scene (a couple of paths being drawn, there are several scenes per slide) I create an own video file and then I also store a png file that contains the last frame of that video.

  I then use kdenlive to join it all together : A video containing the drawing of the first scene, then a png which holds the last image of the video while I talk about the drawing, then the next animated drawing, then the next still image where I continue talking and so on. I use these intermediate images because freezing the last frame is tedious in kdenlive and I have around 600 scenes. Here I do the editing, adjust the duration of the still images and render the final video.

  The background of the video is a photo of a blackboard which never changes, the strokes are paths with a filter to make it look like chalk.

  So far so good, everything almost works.

  My problem is : Whenever there is a transition between an animation and a still image, it is visible in the final result. I have tried several approaches to make this work but nothing is without flaws.

  My first approach was to encode the animations as mp4 like this :

  p = Popen(['ffmpeg', '-y', '-f', 'image2pipe', '-vcodec', 'png', '-r', str(fps), '-i', '-', '-vcodec', 'libx264', '-crf', '21', '-bf', '2', '-flags', '+cgop', '-pix_fmt', 'yuv420p', '-movflags', 'faststart', '-r', str(fps), videofile], stdin=PIPE)

  which is recommended for YouTube. But then there is a little brightness difference between video and still image.

  Then I tried mov with png codec :

  p = Popen(['ffmpeg', '-y', '-f', 'image2pipe', '-vcodec', 'png', '-r', str(fps), '-i', '-', '-vcodec', 'png', '-r', str(fps), videofile], stdin=PIPE)

  I think this encodes every frame as png in the video. It creates way bigger files since every frame is encoded separately. But it’s ok since I can use transparency for the background and just store the chalk strokes. However, sometimes I want to swipe parts of the chalk on a slide away, which I do by drawing background over it. Which would work if those overlaid, animated background chunks which are stored in the video looked exactly like the underlying png in the background. But it doesn’t. It’s slightly more blurry and I believe the color changes a tiny bit as well. Which I don’t understand since I thought the video just stores a sequence of pngs... Is there some quality setting that I miss here ?

  Then I read about ProRes4444 and tried that :

  p = Popen(['ffmpeg', '-y', '-f', 'image2pipe', '-vcodec', 'png', '-r', str(fps), '-i', '-', '-c:v', 'prores_ks', '-pix_fmt', 'yuva444p10le', '-alpha_bits', '8', '-profile:v', '4444', '-r', str(fps), videofile], stdin=PIPE)

  and this actually seems to work. However, the animation files become larger than the bunch of png files they contain, probably because this format stores 12 bit per channel. This is not thaat horrible since only intermediate videos grow big, the final result is still ok.

  But ideally there would be a lossless codec which stores in rgb colorspace with 8 bit per channel, 8 bit for alpha and considers only the difference to the previous frame (because all that changes from frame to frame is a tiny bit of chalk drawing). Is there such a thing ? Alternatively, I’d also be ok without transparency but then I have to store the background in every scene. But if only changes are stored from frame to frame within one scene, that should be manageable.

  Or should I make some fundamental changes in my workflow altogether ?

  Sorry that this is rather lengthy, I appreciate any help.

  Cheers !