[MediaConvert] The video is blurry when using MediaConvert input clip to create a sample video

Question

I have an issue when using MediaConvert input clip to create a sample video.

The video I get after conversion is blurrier than the original video. It only happens when I choose "Start timecode" other than 00:00:00:00, for example: 00:00:01:00

Original Video: [https://s3-07-24.s3.ap-southeast-1.amazonaws.com/original-video.mp4](https://s3-07-24.s3.ap-southeast-1.amazonaws.com/original-video.mp4)

Sample Video: [https://s3-07-24.s3.ap-southeast-1.amazonaws.com/sample-video.mp4](https://s3-07-24.s3.ap-southeast-1.amazonaws.com/sample-video.mp4)

This is my setting:

```
{
  "Queue": "arn:aws:mediaconvert:ap-southeast-1:311311747803:queues/Default",
  "UserMetadata": {},
  "Role": "arn:aws:iam::311311747803:role/service-role/MediaConvert_Default_Role",
  "Settings": {
    "TimecodeConfig": {
      "Source": "ZEROBASED"
    },
    "OutputGroups": [
      {
        "Name": "File Group",
        "Outputs": [
          {
            "ContainerSettings": {
              "Container": "MP4",
              "Mp4Settings": {}
            },
            "VideoDescription": {
              "ScalingBehavior": "DEFAULT",
              "TimecodeInsertion": "DISABLED",
              "AntiAlias": "ENABLED",
              "Sharpness": 50,
              "CodecSettings": {
                "Codec": "H_264",
                "H264Settings": {
                  "MaxBitrate": 5000000,
                  "RateControlMode": "QVBR",
                  "SceneChangeDetect": "TRANSITION_DETECTION"
                }
              },
              "AfdSignaling": "NONE",
              "DropFrameTimecode": "ENABLED",
              "RespondToAfd": "NONE",
              "ColorMetadata": "INSERT"
            },
            "Extension": "mp4"
          }
        ],
        "OutputGroupSettings": {
          "Type": "FILE_GROUP_SETTINGS",
          "FileGroupSettings": {
            "Destination": "s3://s3-07-24/sample-video"
          }
        }
      }
    ],
    "FollowSource": 1,
    "Inputs": [
      {
        "InputClippings": [
          {
            "EndTimecode": "00:00:09:00",
            "StartTimecode": "00:00:01:00"
          }
        ],
        "AudioSelectors": {
          "Audio Selector 1": {
            "DefaultSelection": "DEFAULT"
          }
        },
        "VideoSelector": {
          "ColorSpace": "FOLLOW",
          "SampleRange": "FOLLOW",
          "Rotate": "AUTO"
        },
        "FilterStrength": 5,
        "TimecodeSource": "ZEROBASED",
        "FileInput": "s3://s3-07-24/original-video.mp4"
      }
    ]
  },
  "BillingTagsSource": "JOB",
  "AccelerationSettings": {
    "Mode": "DISABLED"
  },
  "StatusUpdateInterval": "SECONDS_60",
  "Priority": 0
}
```

Can you explain how I can fix it?

Answer

AWS MediaConvert, when creating clips, will look for the nearest I-Frame/IDR in the source file, before the clip Start Time Code. It will then decode from the IDR to and beyond the Start Time code.

The source example that was included seems to be fragmented MP4, with a fragment for each frame, which is a bit unusual. Typically, fragments contain multiple frames, up to several secs in duration. Currently MediaConvert will look for IDRs within a fragment or segment of the source file. The one frame per fragment is limiting MediaConvert from finding an IDR for a decode start. This means it starts its decode with a P or B frame and thus not a full image frame. This creates the blurring experienced until an I-Frame is encountered.

A possible work around is to have your sourced created with fragments that contain 1-2 seconds of video. If this is not achievable, locate where I-Frames are in the source and start the clipping at a fragment that contains an I-Frame.

[MediaConvert] The video is blurry when using MediaConvert input clip to create a sample video

相關內容