Recherche avancée

Sur d’autres sites (6188)

• Ffmpeg : 4K RGB->YUV realtime conversion

  6 mars 2021, par andersd

  I'm trying to use Ffmpeg for creating a hevc realtime stream from a Decklink input. The goal is high quality HDR stream usage with 10 bits.
The Decklink SDI input is fed RGB 10 bits, which is well handled by ffmpeg with the decklink option -raw_format rgb10, which gets recognized by ffmpeg as 'gbrp10le'.

  


  I have a Nvidia pascal-based card, which supports yuv444 10 bit (as 'yuv444p16le') and when when using '-c:v hevc_nvenc' the auto_scaler kicks in and converts to 'yuv444p16le', which I guess is the same conversion as giving '-pix_fmt yuv444p16le'.

  


  This is working very well in 1920x1080 resolution, but in 4096x2160 resolution ffmpeg can't keep up realtime 24 or 25 fps, and I get input buffer overruns.
The culprit seems to be the RGB->YUV conversion in ffmpeg swscale because ;

  


  

  • When piping the Decklink 4K RGB input with '-c:v copy' straight to /dev/null, there's is no problems with buffer underruns,

  


  • And when feeding the Decklink YUV and giving '-raw_format yuv422p10’ (no YUV444 input for decklink seems available for decklink in ffmpeg) I get no underrun and everything works well in 4K. Even if I set '-pix_fmt yuv444p16le'.

  


  


  Any ideas how I could accomplish a 4K hevc in NVENC with the 10-bit RGB signal from the Decklink ? Is there a way to make NVENC accept and use the RGB data without first converting to YUV ? Or is there maybe a way to convert gbrp10le->yuv444p16le with cuda or scale_npp filter ? I have compiled ffmpeg with npp and cuda, but I cannot figure out if I can get it to work with RGB ? Whenever I try to do '-vf "hwupload_cuda"', auto_scaler kicks in and tries to convert to yuv on the cpu, which again creates underruns.

  


  Another thing I guess could help is if there was a way to make the swscale cpu filter(or if there is another suitable filter ?) use multiple threads ? Right now it seems to only use one thread at a time, maxing out at 99% on my Ryzen 3950x (3,5GHz, 32 threads).

  


  Example ffmpeg output :

  


  $ ffmpeg -loglevel verbose -f decklink -raw_format rgb10 -i "Blackmagic Card 1" -c:v hevc_nvenc -preset medium -profile:v main10 -cbr 1 -b:v 20M -f nut - > /dev/null
--
Stream #0:1: Video: r210, 1 reference frame, gbrp10le(progressive), 4096x2160, 6635520 kb/s, 25 tbr, 1000k tbn, 1000k tbc
--
[graph 0 input from stream 0:1 @ 0x4166180] w:4096 h:2160 pixfmt:gbrp10le tb:1/1000000 fr:25000/1000 sar:0/1
[auto_scaler_0 @ 0x4168480] w:iw h:ih flags:'bicubic' interl:0
[format @ 0x4166080] auto-inserting filter 'auto_scaler_0' between the filter 'Parsed_null_0' and the filter 'format'
[auto_scaler_0 @ 0x4168480] w:4096 h:2160 fmt:gbrp10le sar:0/1 -> w:4096 h:2160 fmt:yuv444p16le sar:0/1 flags:0x4
[hevc_nvenc @ 0x4139640] Loaded Nvenc version 11.0
--
Stream #0:0: Video: hevc (Rext), 1 reference frame (HEVC / 0x43564548), yuv444p16le(tv, progressive), 4096x2160 (0x0), q=2-31, 2000 kb/s, 25 fps, 51200 tbn
--
[decklink @ 0x40f0900] Decklink input buffer overrun!:02.52 bitrate= 30471.3kbits/s speed=0.627x


  


• ffmpeg best parameteres for fastest compression on FullHD mp4 ?

  3 août 2021, par Mike Azatov

  I need to compress a bunch of FullHD(1920x1080, 3Gb, mp4) videos, each 2 hrs long. Some loss of quality is okay as is the lower resolution. I'm shooting for a 500Mb output video in 10 min processing time (2080TI Ryzen 9 3900X). It is important for the videos to be completely in sync, so frame x in the input video matches frame x in output. For this, I'm using -vsync 0

  


  My current working settings which get me to 500MB size in 20 minutes.

  


  ffmpeg -i input_fullhd.mp4 -vcodec libx265 -crf 28 -vsync 0 -vf scale=800:-1 -preset medium output.mp4

  


  I'd like it to be faster with potentially slightly better quality.

  


  My concern is that I'm not utilizing my GPU at all and my CPU shows only 30% utilization. So feels like it should be possible to speed this up. I tried adding -hwaccel cuda but that made things a bit slower.

  


  Are there any other settings I can try that I'm missing to either improve my conversion or utilize more resources for speed ?

  


  Thanks

  


• what is the faster way to load a local image using javascript and / or nodejs and faster way to getImageData ?

  4 octobre 2020, par Tom Lecoz

  I'm working on a video-editing-tool online for a large audience.
Users can create some "scenes" with multiple images, videos, text and sound , add a transition between 2 scenes, add some special effects, etc...

  


  When the users are happy with what they made, they can download the result as a mp4 file with a desired resolution and framerate. Let's say full-hd-60fps for example (it can be bigger).

  


  I'm using nodejs & ffmpeg to build the mp4 from HtmlCanvasElement.
Because it's impossible to seek perfectly frame-by-frame with a HtmlVideoElement, I start to convert the videos from each "scene" in a sequence of png using ffmpeg.
Then, I read my scene frame by frame and , if there are some videos, I replace the videoElements by an image containing the right frame. Once every images are loaded, I launch the capture and go to the next frame.

  


  Everythings works as expected but it's too slow !
Even with a powerfull computer (ryzen 3900X, rtx 2080 super, 32 gb of ram , nvme 970 evo plus) , in the best case, I can capture basic full-hd movie (if it contains videos inside) at 40 FPS.

  


  It may sounds good enought but it's not.
Our company produce thousands of mp4 every day.
A slow encoding process means more servers at works so it will be more expensive for us.

  


  Until now, my company used (and is still using) a tool based on Adobe Flash because the whole video-editing-tool was made with Flash. I was (and am) in charge to translate the whole thing into HTML. I reproduced every feature one by one during 4 years (it's by far my biggest project) and this is the very last step but even if the html-version of our player works very well, the encoding process is much slower than the flash version - able to encode full-hd at 90-100FPS - )

  


  I put console.log everywhere in order to find what makes the encoding so slow and there are 2 bottlenecks :

  


  As I said before, for each frame, if there are videos on the current scene, I replace video-elements by images representing the right frame at the right time. Since I'm using local files, I expected a loading time almost synchronous. It's not the case at all, it required more than 10 ms in most cases.

  


  So my first question is "what is the fastest way to handle local image loading with javascript used as final output ?".

  


  I don't care about the technology involved, I have no preference, I just want to be able to load my local image faster than what I get for now.

  


  The second bottleneck is weird and to be honest I don't understand what's happening here.

  


  When the current frame is ready to be captured, I need to get it's data using CanvasRenderingContext2D.getImageData in order to send it to ffmpeg and this particular step is very slow.

  


  This single line

  


  let imageData = canvas.getContext("2d").getImageData(0,0,1920,1080);  


  


  takes something like 12-13 ms.
It's very slow !

  


  So I'm also searching another way to extract the pixels-data from my canvas.

  


  Few days ago, I found an alternative to getImageData using the new class called VideoFrame that has been created to be used with the classes VideoEncoder & VideoDecoder that will come in Chrome 86.
You can do something like that

  


  let buffers:Uint8Array[] = [];
createImageBitmap(canvas).then((bmp)=>{
   let videoFrame = new VideoFrame(bmp);
   for(let i = 0;i<3;i++){
      buffers[i] = new Uint8Array(videoFrame.planes[id].length);
      videoFrame.planes[id].readInto(buffers[i])
   }
})


  


  It allow me to grab the pixel data around 25% quickly than getImageData but as you can see, I don't get a single RGBA buffer but 3 weirds buffers matching with I420 format.

  


  In an ideal way, I would like to send it directly to ffmpeg but I don't know how to deals with these 3 buffers (i have no experience with I420 format) .

  


  I'm not sure at all the solution that involve VideoFrame is a good one. If you know a faster way to transfer the data from a canvas to ffmpeg, please tell me.

  


  Thanks for reading this very long post.
Any help would be very appreciated