Recherche avancée

Sur d’autres sites (7112)

• Upload a picture to generate a video with special effects

  8 novembre 2019, par Zedd

  I am stucked by a video processing feature, Specifically, upload an image and then generate a video based on various video templates.

  Here are the video templates :

  1. http://video-static.biku8.com/data/video/template/3286012076458048/7437ab55-2e83-4a36-9046-5708fcddf4c1.mp4
  2. http://video-static.biku8.com/data/video/template/3274256089907264/ae8fa3f7-6c9c-45ca-810f-48db92cc14cb.mp4
  3. http://video-static.biku8.com/data/video/template/3213894231425088/bf107d439b9043a58c1ea0ba26f811db_template.mp4
  4. ...

  As shown in the video templates above, I just need to upload a photo to generate a great video.

  My question

  1. What is the specific idea for implementing this video ?
  2. Which third-party libraries are needed ? (ffmpeg, opencv)

  PS : I am using dlib and opencv for face recognition. I can generate face image, but I don’t know how to insert face image into the correct position of these template videos.

• FFMPEG crop a portrait image square with respect to the zoomed dimensions and x/y pan

  3 octobre 2022, par hugger

  I am making a photo crop component for my mobile app.

  


  If an image is selected from the image picker and it is above 4:5 ratio, using FFMPEG I need to crop this image 1:1 using the dynamic X/Y values along with the scale value from pinching (1.00 - 10).

  


  I trying to use these values with the FFMPEG crop/scale filter, but no matter what I try I cant seem to get the crop to work as expected... It is not matching up from the UI pan / zoom.

  


  My FFMPEG command looks like this, along with some other relevant variable code :

  


  let zoom = this._scale.__getValue(); //set dynamically by the user; ranges from 0.1 to 10
let translateY = this._translateX.__getValue(); //amount from left side
let translateX = this._translateY.__getValue(); // amount from top

//trying to first crop the width and height / the zoom for the zoom scaling..?
//then I am using translate X/Y to get my coordinates (not matching up)
//I tried to use translateXY * zoom to get the scaling factor but it still did not work...
//the image needs to be square, chain scale filter to make this happen after the original crop?)
FFmpegKit.execute(
  `-y -i ${this.state.mediaSource} -vf "crop=iw/${zoom}:ih/${zoom}:${translateX}:${translateY}, scale=iw:iw:0:0" -qscale 0 -frames:v 1 ${filterPathPostCrop}`
).then(async (session) => {
  const returnCode = await session.getReturnCode();
  if (ReturnCode.isSuccess(returnCode)) {
    this.setState({
      mediaSource: filterPathPostCrop,
      videoSourcePreview: `${filterPathPostCrop}?${new Date().getTime()}`,
      ffMPEGinProgress: null,
      aspectRatio: 1080 / 1080,
      videoTime: 0,
      isPlayingVideo: false,
      isCropping: false,
      filterOutputIsAlt: !this.state.filterOutputIsAlt,
      wasCropped: true,
    });
  } else if (ReturnCode.isCancel(returnCode)) {
    // CANCEL
  } else {
    // ERROR
    alert('error');
  }
});


  


  I appreciate any guidance I can get with this, I feel like I am close I just cant seem to get this calculation working...

  


• What is the ideal image format extraction for YUV color space and lossless compression from mxf video format using ffmpeg

  15 février 2016, par DragonDance27

  I would like to be able to work on frames from an .mxf video file. The image format of the video is JPEG 2000, using a YUV color space with 4:2:2 subsampling and lossless compression.

  My intentions are to extract frames from this video using ffmpeg. Extracted frames would then be processed in Matlab (at the moment I’m interested in performing colorization).

  I want to extract the frames with as minimal data loss as possible and I would like to work in the YUV color space. I understand PNG involves a lossless process, but only involves the RGB color space - so not an option.

  I think I can extract jpeg2000 frames in the YUV color space, but I’m not sure if I’m losing data from the compression process. I attempted the following code in ffmpeg :

  ffmpeg -i video.mxf -r 1/5 out%03d.jp2

  ... however, the extracted jp2 files are unreadable in various software, including HiView which is a specialised JPEG 2000 software.

  Quesiton 1 : Is this jpeg 2000 extraction method lossless ? What am I doing wrong ?

  I also considered extracting the images in the tiff format where I can achieve the YUV and lossless requirements. I attempted the following code in ffmpeg :

  ffmpeg -i video.mxf -vcodec tiff f%10d.tif

  ... however, the extracted tiff files are unreadable in software such as paint, paint.net and windows photo viewer.

  Question 2 : Is this tiff extraction method correct ? What am I doing wrong ?

  Question 3 : What is an ideal image format that covers my YUV and lossless requirements ?