Recherche avancée

Sur d’autres sites (9206)

• Movie making from lyrics with timestamps in python

  26 mai 2020, par carl

  I have lyrics from musixmatch with timestamps. I want to form video with the lyric lines in the video along with images in a file numbered from 1-n.

  



  As seen in this post, I thought ffmpeg would be something which can help me but there aren't much information that i can find.

  



  Also the answer given by @llogan, gives very vague idea of forming videos with "subtitle filter" in ffmpeg.

  



  It would be very helpful you can provide an example to explain your idea.(It can also help other coders at any time :) )

  



  Thanks in advance

  


• Why the source order matters on working with multiple media sources in a single AVFormatContext ?

  24 avril 2024, par Mehmet YILMAZ

  avformat_open_input() deletes the AVFormatContext* and returns -6 when the source order changes.

  


  I am trying to open multiple media sources dynamically with different(mixed) formats and codecs in a single context (AVFormatContext).

  


  My media sources are a BlackMagic DeckLink Duo SDI input as first source and an mp4 file or rtsp stream as second.

  


  When I order to open (avformat_open_input()) the source 2 (RTSP or MP4 file) at first and then open the BlackMagic DeckLink Duo, proceed as expected.

  


  But when I change the order, and first open the DeckLink and then try to open RTSP stream or MP4 file, as I inspected in the step debugger ; AVFormatContext* deleting in the av_open_input() function and it returns -6 as result.

  


  Please find the simple error reproduction code snappet below ;

  


  AVFormatContext* context{avformat_alloc_context()};
const char* url_source1{"DeckLink Duo (1)"};
const AVInputFormat* format_source1{av_find_input_format("decklink")};

const char* url_source2{"http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/BigBuckBunny.mp4"};

// Open the first media input
int result = avformat_open_input(&context, url_source1, format_source1, NULL);

if(result < 0) {
  exit(1);
}

// Open the second media input
// This function in current order deletes the context and returns -6
result = avformat_open_input(&context, url_source2, NULL, NULL);
if(result < 0) {
  exit(1);
}

// Since the context has been deleted in previous step, segmentation fault accours here!
result = avformat_find_stream_info(context, NULL);
if(result < 0) {
  exit(1);
}

std::cout << "Total number of streams: " << context->nb_streams << std::endl;



  


  But When I change the order and call the avformat_open_input() first for the mp4 file and then the DeckLink device as following it proceed as expected, no error.

  


  AVFormatContext* context{avformat_alloc_context()};
const char* url_source1{"DeckLink Duo (1)"};
const AVInputFormat* format_source1{av_find_input_format("decklink")};

const char* url_source2{"http://commondatastorage.googleapis.com/gtv-videos-bucket/sample/BigBuckBunny.mp4"};


// Open the second media input
int result = avformat_open_input(&context, url_source2, NULL, NULL);
if(result < 0) {
  exit(1);
}


// Open the first media input
result = avformat_open_input(&context, url_source1, format_source1, NULL);

if(result < 0) {
  exit(1);
}


result = avformat_find_stream_info(context, NULL);
if(result < 0) {
  exit(1);
}

std::cout << "Total number of streams: " << context->nb_streams << std::endl;



  


• Adding slides of a pdf presentation to the audio file of the same presentation with ffmpeg

  27 mars 2024, par revher

  I want to add slides of a pdf presentation to the audio file of the same presentation with ffmpeg (or whatever).

  


  I split the pdf presentations using pdftoppm into numbered png images and tried to use ffmpeg to sync the audio file with the images. But I am wondering if the output file could be smaller.
Example, the following command will split file 'foo.pdf' into file 'output-01.png', 'output-02.png' etc :

  


  $ pdftoppm foo.pdf -png output


  


  Then I listen to the audio file 'foo-audio.mp3' and noted the end of each slide in minutes and second. Using a spreadsheet you have to calculate the exact (in seconds) timing of each slide, which can be expressed as 40 for 40 seconds for the program of the conference or 00:01:37 for 1 minute and 37 seconds for your first slide etc and paste it into a text file 'foo-images.txt' which will look like :

  


  file ./program.png
duration 40 
file ./output-01.png
duration 00:01:37


  


  Then the video can be built with the command :

  


  $ ffmpeg -safe 0 -f concat -i foo-images.txt -fps_mode vfr -pix_fmt yuv420p  output.mp4
 height not divisible by 2 (1500x1125)type here


  


  but I got the above error that the height was not divisible by 2.
Then I found (on a the ffmpeg mailing list) that I could use the '-vf scale=1680 :-2' option and the following command was successful :

  


  $ ffmpeg -safe 0 -f concat -i foo-images.txt -vf scale=1680:-2  -pix_fmt yuv420p  output.mp4



  


  and in order to sync the audio with the images and get the final video 'foo-video.mp4' :

  


  $ ffmpeg -safe 0 -f concat -i foo-images.txt -i foo-audio.mp3 -vf scale=1680:-2  -pix_fmt yuv420p  foo-video.mp4      



  


  It worked, but the video is about 60Mb for 30 slides and 30 minutes presentation. Is is a correct ratio ? Although the video can be viewed by an Apple viewer. But is there a benefit to use ffmpeg and command lines compared to the use of softwares like imovie or acrobat to get the same result ?
Also I am confused with the scale 1680 (which is fine) and the 420p of pix_fmt.