Posted on October 28, 2020
By Craig Seidel
File naming is a mundane yet persistently exasperating problem across
media production...
In this post, I’m going to explain how we have addressed the problem in
VFX, and how we plan to expand our solution to other areas of the
workflow.
Earlier this year MovieLabs partnered with the Entertainment
Technology Center at USC (ETC) with the idea of developing a focused
specification on “plate” naming. We published a paper in August called VFX
Image Sequence Naming which defined naming for frame-based video (e.g.,
plates and comps) that are inputs and outputs of the VFX process. Now
MovieLabs is expanding this scope to create a more universal
solution.
Files, files, and more files.
The professional media creation process requires a staggering number of
files. Every second a camera is running it generates about 24 files. These
files can be copied every time there is a change, and often duplicated for
other reasons. A major production might create 200,000 image files per
day.
Each specialty creates its own files: Artwork, spreadsheets, computer
models (both the VFX kind and the financial kind), call sheets, timecards,
metadata, continuity, and on and on… These files are shared, often between
disparate file management systems which often rename or recode them.
Keeping files organized is essential to the creative process.
People have developed numerous conventions for naming files so they are
as self-identifiable as possible. Each of these many schemes is tailored
to a specific file type, task, or even to a specific person working on a
task.
That’s fine until files need to be shared, stored or archived. Then the
sender and the receiver must come to some form of agreement and the file
organization must be transformed from one to the other.
Some translations are built into tools. Some require custom work for
bespoke production workflows (i.e., each new production), and some require
manual manipulation.
Our goal is to provide naming conventions that increase interoperability,
reduce custom work, reduce errors, and provide more flexibility in the
creative process—all while keeping implementation as simple as possible to
drive our ultimate goal of enabling creative teams to spend more time
creating and less time managing files.
The fastest way to do something is not to do it!
That means the fastest way to convert files is not to convert them. The
Image Sequence file naming specification we published provides naming that
requires no translation (assuming everyone complies). In other words, a
production can perform a VFX pull from editorial and deliver those files
to any VFX house in exactly the form they expect them. No work. No errors.
Plenty of flexibility.
This project started as “Plate” file naming.
Loosely defined, a Plate is a frame or series of frames used in visual
effects. The term allegedly comes from painted matte plates. Plates are
backgrounds, foregrounds, effects (non-digital or “practical”), or other
images captured where pixels from the image make their way into the final
work. Productions might shoot a “clean” plate without actors to provide
unobstructed background for replacing wires used to suspend actors (rig
removal). Plates can be thought of as video (not audio), but in practice
each frame is its own image file.
When a plate is passed to a VFX team, the recipient must understand the
context of the plate. What is it part of? How does it align with other
plates, and other elements? What work is required on the plate? Our
challenge was to distill the information that describes a plate into a
concise file name that is sufficient to identify the plate.
Aha! It’s an identification problem.
Realizing this was essential to determining what goes in the filename. We
only need enough information to distinguish this plate from all other
plates, thereby establishing its own identity. In other words,
characteristics that aren’t distinguishing (e.g., pixel aspect ratio) need
not be in the filename.
The anatomy of a filename.
Working with the subject matter experts participating in the ETC working
group, we sorted through all the options and narrowed the choices down.
Here are the fields we chose for the final specification:
<showid> + “_” +
<vfx-sequence> + “_” +
<vfx-shot> + “_” +
<image-type> + “_” +
<vendor-code> + “_” +
<revision-code>
{ + “_” + <alternate>}
{ + “_” + <camera-reference>}
{ + “_” + <identifying-description>}
{ + “_” + <spec-version>}
{ + “.” + <frame-number>}
+ “.” + <file-extension>
Show ID, VFX Sequence, and VFX Shot represent Use Context (how it will be
used). Image Type is the kind of frame it is (e.g., main plate), based on
a concept we call Functional Class. There are fields for vendor
identification, revision, variant (e.g., 4K vs 6K), camera, and frame
number. We use file extension to identify its Structural Type, represented
as file type (generally, OpenEXR or DPX). For each of these we defined
encoding rules that would ensure that filenames were encoded
consistently.
Aha #2, It’s not just about plates.
Our second epiphany was that with a few changes we could generalize the
specification to address almost any image sequences in VFX (e.g., comps,
concept, layout, look dev, test shots, etc.). A few tweaks and the “Plate
Naming” spec became the “Image Sequence Naming” spec. That is what we
published.
Thanks!
Special thanks to the ETC working group co-chairs Erik Weaver and Horst
Sarubin of Universal Studios. Thanks also to the contributors listed on
page iii of the spec. Notably, Horst laid the foundation for this spec,
and Barbara Ford Grant of MovieLabs provided domain expertise in honing
it.
What’s next?
We are extending the work to other VFX focused areas (e.g., still
images). This is part of a broader MovieLabs effort to establish a
framework for naming any asset throughout the production process, to help
increase interoperability, reduce errors, and save time.
As part of
our work to realize the MovieLabs 2030 Vision, we have developed an asset
metadata model that we are using to structure the metadata elements that
comprise a filename. But that’s a topic for a future blog.
In the meantime, please download the specification and let us know if you
have any comments. We’re interested in your feedback and participation on
both the general naming and VFX naming work—because ultimately the
benefits of standardization are only realized when everyone uses the
standard.
