Recherche avancée

Sur d’autres sites (9675)

• swscale/x86/output : add AVX2 version of yuv2nv12cX

  26 avril 2020, par Nelson Gomez

  swscale/x86/output : add AVX2 version of yuv2nv12cX
  256 bits is just wide enough to fit all the operands needed to vectorize
  
  the software implementation, but AVX2 is needed to for a couple of
  
  instructions like cross-lane permutation.
  Output is bit-for-bit identical to C.
  Signed-off-by : Nelson Gomez <nelson.gomez@microsoft.com>
  

  • [DH] libswscale/x86/output.asm
  • [DH] libswscale/x86/swscale.c

• Are there people interested in converting ffmpeg source to Go ?

  30 septembre 2018, par No One

  After seeing that Go compiler have been converted from C to Go I thought same for ffmpeg ? Don’t want to go deep into reasons as I think they are obvious. It was very hard to be so close to the have rich library as ffmpeg in other language. It was even hard to make bindings for that scale of library. I’m not enough advanced to start something like this myself, so is there anybody else interested in this ? If yes then where this question worth to be addressed, so people interested in this may have discussion ?

  (Seems not enough obvious so adding some details.)

  For applications which use large amount of commands with different complexity it is hard to read the code as it’s not actually a code. Instead, it’s commands which you will need to understand by reading docs from ffmpeg’s docs page. I had used ffmpeg before in Nodejs and there was lots of logic of manipulating command string. Also sometimes in windows it was ending with cmd limitations error. When you are working with some language it is nice to see whole logic in that language. So you know go ? than you know everything that is happening with this code without even going off from code and reading docs of another application.

  There may be some benefits from executing stuff in goroutines so you can handle concurrency in the way you want not in the way it is implemented in ffmpeg.

  Build faster with Go.

  Less code.

  Possibility to split code into smaller packages.

  Also if you are familiar why community converted compiler from C to Go than I think some reasons will fit too.

• FFMPEG on AWS Lambda works selectively

  20 novembre 2020, par Arindam Baral

  I am using FFMpeg on AWS Lambda to convert mp4 files to m3u8 format.

  


  Here's my code :

  


  ENCODE_SIZE = {
    '360p': ('360:640', '1000000'),
    '540p': ('540:960', '2000000'),
    '720p': ('720:1280', '5000000')
}
@app.route('/encode', methods=['GET'])
@token_required
def encodeVideo(current_user):
    try:
        userEmail = current_user.emailID
        
        courseID = request.args.get('courseID')
        fileKey = request.args.get('fileKey')
        print ('Input file key received ', fileKey)
        convType = 'all'
        if 'convType' in request.args:
            convType = request.args.get('convType')
        print ('Conv Type::::', convType)
        instiID = Course.query.filter_by(id=courseID).first().instiID
        adminEmail = Institute.query.filter_by(id=instiID).first().adminEmail
        if adminEmail != userEmail:
            return jsonify({'message': 'Not authorized'}), 403
        
        bucket = app.config['S3_CONTENT_FOLDER']
        folder = '/'.join(fileKey.split('/')[:-1])
        
        os.system('cp /var/task/ffmpeg /tmp/; chmod 755 /tmp/ffmpeg;')
        FFMPEG_STATIC = '/tmp/ffmpeg' #"/opt/bin/ffmpeg" # 
        # FFMPEG_STATIC = 'ffmpeg' #"/opt/bin/ffmpeg" # 
        
        preSignURL = getPreSignedS3URL(fileKey, bucket)
        print (preSignURL)
        
        outputFlag = []
        
        # outFileName = 'final_out.m3u8'
        outFileName = '/tmp/final_out.m3u8'
        with open(outFileName, 'a') as outFile:
            outFile.write('#EXTM3U\n')
            
            if convType == 'all':
                for ver in ENCODE_SIZE:
                    print ('Starting for ver ', ver)
                    outFileNameM3 = '/tmp/%s_out.m3u8' %(ver) 
                    # outFileNameM3 = '%s_out.m3u8' %(ver) 
                    
                    subprocess.call([FFMPEG_STATIC, '-i', preSignURL, '-c:a', 'aac', '-c:v', 'libx264', '-s', ENCODE_SIZE[ver][0], '-f', 'hls', '-hls_list_size', '0', '-hls_time', '10', outFileNameM3])
                    
                    outFile.write('#EXT-X-STREAM-INF:PROGRAM-ID=1, BANDWIDTH=%s %s\n' %(
                        ENCODE_SIZE[ver][1], outFileName
                    ))
                    # ret = os.system(commandStr)
                    # outputFlag.append(ret)
                    print ('Encoding completed for ver ', ver)
            else:
                ver = convType
                outFileNameM3 = '/tmp/%s_out.m3u8' %(ver) 
                # outFileNameM3 = '%s_out.m3u8' %(ver) 

                subprocess.call([FFMPEG_STATIC, '-i', preSignURL, '-c:a', 'aac', '-c:v', 'libx264', '-s', ENCODE_SIZE[ver][0], '-f', 'hls', '-hls_list_size', '0', '-hls_time', '10', outFileNameM3])  
                    
                outFile.write('#EXT-X-STREAM-INF:PROGRAM-ID=1, BANDWIDTH=%s %s\n' %(
                        ENCODE_SIZE[ver][1], outFileName
                ))
                # ret = os.system(commandStr)
                # outputFlag.append(ret)
        outFile.close()
        print ('File Key Generated')
        #Upload files to s3
        streamFiles = glob.glob('/tmp/*.ts')
        print (streamFiles)
        for fl in streamFiles:
            finFileName = fl.split('/')[-1]
            fileKeyName = folder + '/' + finFileName
            uploadFileToS3(fl, fileKeyName, app.config['S3_CONTENT_FOLDER'])
        # m3u8Files = glob.glob('/tmp/*.m3u8')
        # print (m3u8Files)
        # for fl in m3u8Files:
        #     finFileName = fl.split('/')[-1]
        #     fileKeyName = folder + '/' + finFileName
        #     uploadFileToS3(fl, fileKeyName, app.config['S3_CONTENT_FOLDER'])
        # print ('S3 upload completed')
        
        # files = glob.glob('/tmp/*')
        # for f in files:
        #     os.remove(f)
        return jsonify({'message': 'Completed file encoding'}), 201
    except:
        print (traceback.format_exc())
        return jsonify({'message': 'Error in encoding file'}), 504
if __name__ == '__main__':
    app.run(debug=True,host="0.0.0.0", port=5000)



  


  When I request for "540p", the m3u8 file gets converted perfectly.
However, when I request for "360p" and "720p", I see only the first 1 second of the video.

  


  Please note - All the tls files are generated in all the cases. The problem is only in creation of the m3u8 playlist file.

  


  Can someone please help in this regard ?

  


  EDIT 1 : I am quite new to FFMPEG. So any help here will be very welcome

  


  The log files are present here :
https://drive.google.com/file/d/1QFgi-4jDIN3f6mWLoR999mlXzZ-Ij83R/view?usp=sharing