Archives Digital Video — 22 January 2015

It is common now for a single oral history interview that was recorded on video to span multiple video data files. While I advocate maintaining these individual parts as the “master” files in your archival system, delivery of this interview, in those same component parts, will prove to be a very frustrating user experience. At the Nunn Center we have developed a workflow that stitches these parts together in the creation of a single sub-master or “production master” data file that serves as the source for all derivative versions. The purpose of the sub-master or “production master” is to create an archival-quality version of the interview (equivalent in quality and settings to the original) that contiguously stitches together the parts of an interview and serves as a convenient, single-file production master serving as the source for all derivative versions.

Since we are now creating time-dependent metadata (time-coded transcripts and time correlating indexes) using tools such as OHMS, we want to ensure that there is no shift in time-code when generating future versions, or the synchronized transcript or time-coded metadata will no longer correlate. As I wrote in the OHDA article Is Perfect the Enemy of Good Enough? 
Digital Video Preservation in the Age of Declining Budgets, the Nunn Center, for a variety of reasons, has not yet adopted use of a standard “evergreen” format such as the FADGI recommended lossless JPEG2000 contained in an MXF wrapper. For now, we are putting preservation focus on the born-digital originals and creating the sub-master instantiations using production-quality ubiquitous formats. While this does involve transcoding and the addition of compression for the sub-master or “production master,” the use of codecs such as Apple’s ProRes 422 will minimize loss and maintain the production quality of the original. This is not ideal from a variety of perspectives (ProRes is still not considered a “preservation” format), but as I indicate in the title of the article referenced above, perfect is often the enemy of good enough and ProRes–as an encoding format–wrapped in a .mov has many advantages.  In the article cited above, I reference our use of a “Mezzanine” format.  Our Mezzanine format is derived from the sub-master and will not be addressed here.  For more information on Apple’s ProRes 422, check out the article “All in the (Apple ProRes 422 Video Codec) Family” by Kate Murray in the Library of Congress  digital preservation publication The Signal. 

I am basing this post on the manual stitching process using a video editing platform such as Apple’s Final Cut Pro X.  Once you understand the basics of stitching and transcoding, I would recommend consideration automating your workflow.  There are tools that can automate stitching, but this is usually allocated to high-end and complicated workflows using tools such as FFMPEG or Episode by Telestream.  There are lower-end options for software applications that stitch together parts automatically, often done in conjunction with batch transcoding or re-wrapping (changing the container file but leaving the encoding alone).  One such application that I have had success using a low-cost application called AVI Converter Ultimate, which I used to rewrap AVI files as MOV files.  The lower-end applications tend to be a little finicky and buggy; however, if you properly test and evaluate these applications, some of them work great. I still think it is important to learn the basics before you explore automation.

Creating the Sub-master or “Production Master”

In the Nunn Center workflow, all future archival derivatives will be generated from the stitched sub-master, so it is important that original (or as close to original) quality is preserved. Warning: this will result in larger data files. Often an AVCHD original will be blown up from 30 gigabytes to over 100 gigabytes when converted to an Apple ProRes 422 encoded .mov file. For the sub-master, the Nunn Center workflow requires:

  • Following import of video files into video editing system (such as Final Cut Pro X), drop interview components onto the timeline contiguously. We only edit the parts if absolutely necessary. Usually this involves minor trimming of portions of the video before and after breaks. I do not recommend this, as minor editing will create a discrepancy in time code when compared to your original and should only be done if absolutely necessary.
  • Maintain original resolution. For example, if the original was shot using 1920x1080p, continue to use 1920x1080p throughout the system.
  • Maintain original frame rate. If the original was shot using 29.97 or 24 frames per second (fps), continue that setting throughout the system.
  • Maintain original bitrate (depending on the codec used). Usually this entails selecting “maintain from source” or something to that effect. This will ensure no loss in quality in the transcoding process.
  • Use Multi-pass encoding for higher quality.
  • Render: We create our sub-masters or “production masters” using Apple ProRes 422 HQ using a .mov container. If possible, maintain original codec and wrapper, however, this not usually possible since camera formats tend not to equate as good editing formats.  Using Final Cut Pro X, you render the sub-master/”production master” by utilizing the “Share” feature located in the “File” menu. Select “Master File” option and update the settings appropriately.
  • If original is AVC/H.264/HDV/MP4/MXF encoded files (usually a .mov), we will retain the original file format and codec when creating a sub-master or “production master”.
  • If originals are AVCHD, we will import into Final Cut Pro X (involving the “create archive” function, stitch parts together contiguously, and render as an Apple ProRes 422 HQ.

Creating the “Access” Version from the “Stitched” Sub-Master

The Nunn Center’s “access” versions are always derived from the sub-master. Access versions are designed to upload to video servers. Our streaming server has a 2gb file size limit, so we have optimized our settings to achieve a final product that retains visual quality in a small file size and that is optimized for streaming. When creating our access versions, I recommend:

  • Maintain original resolution (see sub-master workflow above).
  • Maintain original frame rate (see sub-master workflow above).
  • Use Multi-pass encoding for higher quality.
  • Optimize for streaming. If your program allows you to choose a “fast start” option, this will make streaming more efficient.  This will make your access version work better with systems such as YouTube or Vimeo.
  • Usually I use Constant Bitrate (CBR) set around 3,000 kbits/second, which will result in a video file under 2 gigabytes. Variable Bitrate (VBR) is best for achieving efficient video quality (utilizes high bitrate when necessary/ low bitrate when necessary) but use of VBR makes it difficult to ensure a predictable file size. Since this is important in our system, we use CBR for generating our access versions.
  • Render: H.264 encoded (high profile) .mov files. H.264 encoded video has multiple profiles. I have found that use of H.264’s “High” profile is best for maintaining quality in HD video.

The resulting files will be reasonable quality for streaming and achieve file sizes around 2gb. These versions are designed to be streaming versions and do not maintain full production quality. I also assume that the “Access” versions will need to be recreated when standards, bandwidth, and codecs inevitably change down the road.  So be sure to document your settings because you will need to migrate in the not-so-distant future.

Conclusion

Stitching adds an often-cumbersome step to the workflow when created manually.  However, since this is important and will provide greater efficiency in the future, we have embraced the additional step in the creation of a sub-master or “production master.”  Once more, you are adding significant data to your preservation package, but you will not have to recreate a stitched version down the road.  As always, pay attention to your codecs and formats so you can effectively migrate your interviews and avoid obsolescence.

Share

About Author

Douglas A. Boyd

(0) Readers Comments

Leave a Reply

Your email address will not be published. Required fields are marked *