Recherche avancée

Sur d’autres sites (12583)

• Screen Recording Issues using Python and FFMPEG [closed]

  3 juin, par Kevin McDowell

  I am trying to record a walkthrough of a 3D space contained within a Chrome browser session. The goal is to create deterministic output. If I run this command from a VS Code terminal, it records 30 seconds of video from the window that's about 6MB :

  


  ffmpeg -f gdigrab -framerate 30 -i "title=My3D - Google Chrome" -t 30 "My3D - Google Chrome_screenRecording_v0.mp4"


  


  However, if I call it from Python like so :

  


  subprocess.run("ffmpeg -f gdigrab -framerate 30 -i \"title=My3D - Google Chrome\" -t 15 \"My3D - Google Chrome_screenRecording_v0.mp4\"")


  


  The output video is 30 seconds, but while I can see the mouse moving on the screen, the image is all black and the filesize is only 94KB.

  


  This is on a Windows 11 machine with GPU acceleration disabled in the OS as well as in Chrome. I've also tried it with other web pages and those are also recorded black. Any ideas what is causing this or how to record the screen using Python ?

  


  I've got other alternatives working (using Selenium or PyGet to capture screenshots and then using the VideoWriter class from OpenCV to write them to a video file), but that has a limitation of only about 10 FPS and you can't control the recording time like you can with FFMPEG. If you screen capture for 30 seconds, the actual output may only be 12-18 seconds.

  


• How to append an image to a video using OpenCV or FFMPEG or Moviepy or other libraries ?

  19 juillet 2022, par Trần Tiến Văn

  Do you know a library in Python to add a frame image to an existing video ? The result video must have the same quality as the image.

  


  I tried to use OpenCV to add google image : https://www.google.com/search?q=google&sxsrf=ALiCzsZhrdoHnOTmg0We4dxtguCqzma5Jg:1657603343101&source=lnms&tbm=isch&sa=X&ved=2ahUKEwiTh8bTzfL4AhWhplYBHfXNAKwQ_AUoAXoECAIQAw&biw=1492&bih=739&dpr=1.25#imgrc=PRtenhDnrVrfOM

  


  But the quality decreases when the video elongates.

  


  Here is the final result video : https://drive.google.com/file/d/1ArDvoX-kN9H_oLbACk3kU1Cid93SMczC/view?usp=sharing

  


  Here is my code using OpenCV :

  


          image = cv2.imread(path_image)
        height, width, dimensions = image.shape
            
        video = cv2.VideoCapture(path_video)
        
        
        frames = []
        while(True):
            
            ret, frame = video.read()
            
            if ret == True: 
                frames.append(frame)
                # frame = frame.resize(frame, (width, height), fx=0, fy=0, interpolation = cv2.INTER_CUBIC)
                
                # Press S on keyboard 
                # to stop the process
                if cv2.waitKey(1) & 0xFF == ord('s'):
                    break
            # Break the loop
            else:
                break
            
        video2 = cv2.VideoWriter(path_video,cv2.VideoWriter_fourcc('M','J','P','G'), 30, (width, height))
        for frame in frames:
            video2.write(frame)
        video2.write(image)
        video2.release()  # releasing the video generated     
        print("Added {}".format(image_name))


  


  I hope to improve the quality of this video.

  


• Automatically inject 360° 3D metadata into a video with a Bash script

  27 octobre 2019, par Eduardo Perez

  So, I have a Cygwin script I made which concats several videos into a single video with different variations, and I want to inject stereoscopic 3D metadata into the videos through the script so I don’t have to inject each video separately using Google’s injector tool. The videos are all 360° videos with top/bottom 3D and standard stereo audio rather than spacial audio, and in an MP4 container. Is there any way that I can either inject the needed 3D metadata using FFmpeg so I can upload it to YouTube as a 360VR video, or use the source code of Google’s injector tool or some other tool in order to inject the metadata the same way that Google’s injector tool would so it’s supported as a 360° 3D video by YouTube ?

  Also, will the injector tool automatically move the MOV atom to the beginning of the file (if the injector tool is used) or will I still need to use -movflags +faststart in FFmpeg ? The videos are kind of big and apparently using FFmpeg to concat several video files together and copy the stream codecs with -movflags +faststart and then injecting the metadata using Google’s Spherical Media tool is three times longer than just using FFmpeg in the same way but without -movflags +faststart, so if there was a fast way to do this I’d greatly appreciate it.