Recherche avancée

Sur d’autres sites (10171)

• How correctly show video with transparency in Qt with OpenCV + FFMpeg

  11 avril 2022, par TheEnigmist

  I'm trying to show a video with transparency in a Qt6 application using OpenCV + FFMPEG.
Actually those are tool versions :

  


  

  • Win 11

  


  • Qt 6.3.0

  


  • OpenCV 4.5.5 (built with CMake)

  


  • FFMPEG 2022-04-03-git-1291568c98-full_build-www.gyan.dev

  


  


  I've used a base .mov video with transparency as test (link provided below).
First of all I've converted .mov video to .webm video (VP9) and I see in output text that alpha channel remains

  


  


  ffmpeg -i '.\Retro Bars.mov' -c:v libvpx-vp9 -crf 30 -b:v 0 output.webm

  


  


  Input #0, mov,mp4,m4a,3gp,3g2,mj2,
    ...
    Stream #0:0[0x1](eng): Video: qtrle (rle  / 0x20656C72), argb(progressive),
    ...

Output #0, webm, 
   ...
   Stream #0:0(eng): Video: vp9, yuva420p(tv, progressive),
   ...


  


  But when I show info of output file with ffmpeg it loses alpha channel :

  


  


  ffmpeg -i .\output.webm

  


  


  Input #0, matroska,webm,
    ...
    Stream #0:0(eng): Video: vp9 (Profile 0), yuv420p(tv, progressive),
    ...


  


  If I open output.webm with OBS it is shown correctly without a background, as shown in picture :


  


  If I try to open it with OpenCV + FFMPEG it shows a black background under bars, as shown in picture :


  


  This is how I load video in Qt :

  


  cv::VideoCapture capture;
capture.open(filename, cv::CAP_FFMPEG);
capture.set(cv::CAP_PROP_CONVERT_RGB, false); // try forcing load alpha channel
... //in a thread
while (capture.read(frame)) {
    qDebug() << "c" << frame.channels() << "t" <<  frame.type() << "d" <<  frame.depth(); // output: c 3 t 16 d 0
    cv::cvtColor(frame, frame, cv::COLOR_BGR2RGBA); //useless since no alpha channel is detected
    img = QImage(frame.data, frame.cols, frame.rows, QImage::Format_RGBA8888);
    emit processedImage(img); // to show image in a QLabel with QPixmap::fromImage(img)
}


  


  I think the problem is when I load the video with OpenCV, it doens't detect alpha channel, since I can load correctly in other player (obs, html5, etc.)

  


  What I'm wrong with all process to show this video in Qt with transparency ?

  


  EDIT : Added dropbox link with test video + ffmpeg outputs :
sample items

  


• Show progress bar of a ffmpeg video convertion

  13 juin 2022, par stackexchange.com-upvm25mz

  I'm trying to print a progress bar while executing ffmpeg but I'm having trouble getting the total number of frames and the current frame being processed. The error I get is AttributeError: 'NoneType' object has no attribute 'groups'. I've copied the function from this program and for some reason it works there but not here, even though I haven't changed that part.

  


  main.py

  


  pattern_duration = re.compile(
    'duration[ \t\r]?:[ \t\r]?(.+?),[ \t\r]?start', re.IGNORECASE)
pattern_progress = re.compile('time=(.+?)[ \t\r]?bitrate', re.IGNORECASE)

def execute_ffmpeg(self, manager, cmd):
    proc = expect.spawn(cmd, encoding='utf-8')
    self.update_progress_bar(proc, manager)
    self.raise_ffmpeg_error(proc)

def update_progress_bar(self, proc, manager):
    total = self.get_total_frames(proc)
    cont = 0
    pbar = self.initialize_progress_bar(manager)
    try:
        proc.expect(pattern_duration)
        while True:
            progress = self.get_current_frame(proc)
            percent = progress / total * 100
            pbar.update(percent - cont)
            cont = percent
    except expect.EOF:
        pass
    finally:
        if pbar is not None:
            pbar.close()

def raise_ffmpeg_error(self, proc):
    proc.expect(expect.EOF)
    res = proc.before
    res += proc.read()
    exitstatus = proc.wait()
    if exitstatus:
        raise ffmpeg.Error('ffmpeg', '', res)

def initialize_progress_bar(self, manager):
    pbar = None
    pbar = manager.counter(
        total=100,
        desc=self.path.rsplit(os.path.sep, 1)[-1],
        unit='%',
        bar_format=BAR_FMT,
        counter_format=COUNTER_FMT,
        leave=False
    )
    return pbar

def get_total_frames(self, proc):
    return sum(map(lambda x: float(
        x[1])*60**x[0], enumerate(reversed(proc.match.groups()[0].strip().split(':')))))

def get_current_frame(self, proc):
    proc.expect(pattern_progress)
    return sum(map(lambda x: float(
                x[1])*60**x[0], enumerate(reversed(proc.match.groups()[0].strip().split(':')))))


  


• Show video length in HLS player before all TS files are created

  14 novembre 2022, par WilliamTaco

  I have a spring-boot backend which on request (on demand) uses ffmpeg to create a m3u8 playlist with its ts files from a mp4 file. So basically my react frontend requests the index.m3u8 from the backend and if it doesnt already exist it creates it and then start serving it with its ts files. This causes the frontend HLS player to show the length of the video to the combined length of the generated chunks which gets longer as time goes on until its fully there. It totally makes sense but was wondering what the correct way of showing the full length in the player even though its not fully created yet ?

  


  Im using react-hls-player for playing the stream and spring-boot + a java ffmpeg wrapper to transcode the video.

  


  Might be thinking about this the wrong way so feel free to correct me if im in the wrong path !