Recherche avancée

Sur d’autres sites (8010)

• HTML5 Video Compatibility (MP4, WEBM, OGG) in 2021

  19 juillet 2021, par William

  The support of HTML5 video has evolved a lot over the years. I am trying to understand whether the <video></video> element still needs to have three sources : MP4, WEBM, and OGG.

  


  There are a lot of answers throughout StackOverflow with deeply conflicting information - some of which say that you just need MP4 now, others say, MP4 and WEBM are enough, and then finally many say that you need all three (although many of those article are 10 years old).

  


  W3 suggests that either MP4 or WEBM alone would have universal support (Even though I found a 2011 article from Google saying that they would be removing support for MP4/H.264). Wikipedia paints a more complicated picture (as well as listing that Google Chrome does indeed support MP4/H.264). Azure Media services ONLY seems to allow output in MP4, which would suggest to me that MP4 must have widespread compatibility.

  


  Also see Example 1, Example 2, Example 3.

  


  Is there any definitive information on what video types to include in an HTML5 video player to achieve widespread compatibility ?

  


  Background : I am building a Content Management Platform that allows uploading videos. When a new video is uploaded, a conversion process kicks off to convert the video into the required formats. This takes time and CPU/Memory, so if it is possible I would like to convert uploaded videos into as few formats as possible.

  


  p.s. This question HAS been asked before, however, the fundamentals of playing video on the web continually evolve and most of the answers out there have become irrelevant.

  


• ffmpeg libavformat - output method

  2 octobre 2022, par Branqueira

  I have made a website that can display x camera live using rtsp.

  


  The user connects to the url, the javascript hls.js makes a request to the tomcat java server. If this is the first invocation, the code does jni to call my c code which uses ffmpeg.
In the ffmpeg part, the input is the rtsp : // url and the output is the hls + http stream which corresponds to my tomcat server. This allows to put in a buffer if there is x connection to see the ip camera.
In conclusion, there is only one ffmpeg instance running regardless of the number of people calling the html page.

  


  All this works well and for 2 years already.

  


  But I would like to improve the "machin"

  


  I would like the output of ffmpeg to always be in hls but instead of it being http, I want this to invoke a method (for example) that contains the file's buffer and its name.

  


  for that I use

  


  avformat_alloc_output_context2 (& formatOutputCtx, NULL, "hls", output_url);


  


  which allows me to select the hls + http protocol automatically

  


  I looked at the doc but couldn't find a solution using another protocol or doing one custom.

  


  If someone has an idea ? otherwise no big deal :)

  


• ffmpeg resize video then over lay on top of image using nvidia cuda

  16 mars 2021, par user72261

  i have a bunch of old family videos that are 4:3 SD format i am looking to change them to 16:9 and over lay the video on top of a 720p background

  


  i can do this fine using standard ffmpeg and everything works but it takes for ever with over 10 years of videos to convert

  


  i have a desktop pc i use for work with 2070 super installed so i was looking to use that to speed it up a little i tried adding cuda hardware support and i just hit errors

  


  ffmpeg -y -vsync 0 -hwaccel cuda -hwaccel_output_format cuda -loop 1 -i image.jpg -i video.mp4 -filter_complex "[1]hwupload_cuda,scale=1280:720:force_original_aspect_ratio=decrease[inner];[0][inner]overlay_cuda=x=10:y=10:shortest=1[out]" -map "[out]" -map 1:a -c:a aac -map 0:v -c:v  h264_nvenc -b:v 5M -y output.mp4


  


  the errors i keep getting are

  


  decoder->cvdl->cuvidDecodePicture(decoder->decoder, &ctx->pic_params) failed -> CUDA_ERROR_INVALID_IMAGE: device kernel image is invalid