How do I stop AWS Elemental MediaConvert chopping off the quiet part of my audio?

Question

Basically I have the same issue mentioned here: https://repost.aws/questions/QUp6krkiXUSmu6YMln7dnHdw/aws-media-convert-removes-silent-audio-which-causes-the-output-video-to-be-severly-out-of-sync-with-audio. That post doesn't give an answer, claiming it was a internal MediaConvert bug which got fixed.

I have a 66 second long .mka audio file, generated by Twilio's video room recording API. I want to convert it to a format accepted by AWS Transcribe, so I am using AWS Elemental MediaConvert to convert it to mp4 (converting to mp3 would also be fine).

The input file is a recording which is mostly just very quiet background noise, but features speech for the last few seconds. AWS Elemental MediaConvert seems to be chopping off the initial part which doesn't include speech, returning an mp4 which is just 14 seconds long corresponding to the last 14 seconds of the input.

But I want the output file to be the same length as the input file. How do I accomplish this?

Here is my MediaConvert template:

{
  "Name": "Experimental job",
  "Settings": {
    "TimecodeConfig": {
      "Source": "ZEROBASED"
    },
    "OutputGroups": [
      {
        "CustomName": "My audio file group",
        "Name": "File Group",
        "Outputs": [
          {
            "ContainerSettings": {
              "Container": "MP4",
              "Mp4Settings": {}
            },
            "AudioDescriptions": [
              {
                "AudioSourceName": "Audio Selector 1",
                "CodecSettings": {
                  "Codec": "AAC",
                  "AacSettings": {
                    "Bitrate": 96000,
                    "CodingMode": "CODING_MODE_2_0",
                    "SampleRate": 48000
                  }
                }
              }
            ]
          }
        ],
        "OutputGroupSettings": {
          "Type": "FILE_GROUP_SETTINGS",
          "FileGroupSettings": {
            "Destination": "my-output-bucket",
            "DestinationSettings": {
              "S3Settings": {
                "StorageClass": "STANDARD"
              }
            }
          }
        }
      }
    ]
  },
  "AccelerationSettings": {
    "Mode": "DISABLED"
  },
  "StatusUpdateInterval": "SECONDS_60",
  "Priority": 0,
  "HopDestinations": []
}

And my MediaConvert parameters:

{
        "Queue": MY_MEDIA_CONVERT_QUEUE_ARN,
        "JobTemplate": "Experimental job",
        "Role": MY_MEDIA_CONVERT_ROLE_ARN,
        "Settings": {
            "Inputs": [
                {
                "AudioSelectors": {
                    "Audio Selector 1": {
                        "Offset": 1,
                        "DefaultSelection": "NOT_DEFAULT",
                        "ProgramSelection": 1,
                        "SelectorType": "TRACK",
                        "Tracks": [1]
                    }
                },
                
                "VideoSelector": {
                  "ColorSpace": "FOLLOW"
                },

                "FilterEnable": "AUTO",
                "PsiControl": "USE_PSI",
                "FilterStrength": 0,
                "DeblockFilter": "DISABLED",
                "DenoiseFilter": "DISABLED",
                "TimecodeSource": "EMBEDDED",
                "FileInput": "my-input-file"
                }
            ]
        },
        "UserMetadata": "mymetadata"
}

I've tried adding an InputGroup which looks like this to my template, but it doesn't help:

    "Inputs": [
      {
        "AudioSelectors": {
          "Audio Selector 1": {
            "DefaultSelection": "DEFAULT",
            "SelectorType": "TRACK",
            "AudioDurationCorrection": "TRACK"
          }
        },
        "TimecodeSource": "ZEROBASED"
      }
    ]

What should I do differently?

Edit: I found that when using ffmpeg to transcode the file from mka to mp3 on the command line, I had to use the option -af aresample=async=1, as per this answer: https://stackoverflow.com/a/52847587/3310775. Doing this resulted in an output file of the correct length. But what MediaConvert setting does this correspond to?

Edit: I would also accept settings to accomplish this using AWS Elastic Transcoder.

Answer 1

• MediaConvert may remove silent audio tracks from the output video, which can cause the output video to be out of sync with the audio from its audio normalization feature - possible workaround to prevent this behavior set audio normalization algorithm to ITU-R BS.1770-4 and the correction type to MEASURE_ONLY & also the audio normalization feature of MediaConvert supports the ITU-R BS.1770-1, -2, -3, and -4 standard algorithms, and allows the selected algorithm to only produce loudness measurements. It also allows logging loudness levels and storing these logs in S3.

• Modify template to include the audio normalization settings under the AudioDescriptions section

"AudioDescriptions": [
  {
    "AudioSourceName": "Audio Selector 1",
    "CodecSettings": {
      "Codec": "AAC",
      "AacSettings": {
        "Bitrate": 96000,
        "CodingMode": "CODING_MODE_2_0",
        "SampleRate": 48000
      }
    },
    "AudioNormalizationSettings": {
      "Algorithm": "ITU_BS_1770_4",
      "AlgorithmControl": "MEASURE_ONLY",
      "LoudnessLogging": "LOG"
    }
  }
]

• Run like above & check the logs. You should see the input max loudness value for each audio track in the log file. If the value is lower than -50 dB, the track is considered silent and may be removed by MediaConvert.
• If you want to keep the silent audio tracks in your output file, either increase the loudness level of the input file using a tool like ffmpeg, or use another service like AWS Elastic Transcoder to mp4 or mp3.latter does not have the audio normalization feature, so it should preserve the silent audio tracks in your output file.

reference to comments :

elastic Transcoder settings can be tried

• Audio Bitrate: increase the audio bitrate to ensure that there is enough data to preserve the quiet parts of the audio. These can be tried . There is no doc say that they worked . Also take a look here

• Other things to try with Audio that offered by it that might preserve the audio. See which works !