Production Media Management: Transforming Media Workflows by leveraging the Cloud


The VFX Plate Generation & Delivery workflow is started by an Editorial team with an edit decision list timeline file (.edl, .xml) exported from a Non Linear Editing tool. This timeline file contains only the references to media with additional information about time, color, markers and more, but not any of the actual media files. In addition to the timeline, the Editor chooses whether they would want the resulting media to be rescaled to UHD and how many extra frames they would like to have added for each event referenced in the timeline.

After processing the timeline file, each individual media reference is extracted with relevant timecode, media reference, color decisions and markers. To support different editorial tools, each having its own edit decision list timeline format, our Video Encoding platform interprets the timeline into a standardized interchange format called OpenTimelineIO.

Media reference, color decisions and markers are linked with the original camera media and transcoded from raw camera formats onto ACES EXR. Most Visual Effects tools are not able to process raw camera files directly. Along with image media, color metadata is extracted from the timeline to generate Color Decision List files (.cdl, .xml) which are used to communicate color decisions made by an Editor. All of the media transformations and metadata are then persisted as VFX Plate assets.

The Editor then reviews VFX Plate Generation & Delivery details, with all of the timeline events clearly identified and any inconsistencies spotted, such as if raw camera media is not found or there are any challenges with transcoding media. If all looks good, an Editor is able to submit this workflow onto the final step where results are packaged and shared with the Visual Effects team.

To share results with Visual Effects artists, we’re transforming all of the VFX Plate assets and media created earlier and sharing with the recipients, who can either download the files via browser, or use our auto-downloader tools for additional convenience. Concluding this workflow is an email, sharing all the relevant information with the Editorial and VFX teams.

We’re leveraging the VFX Plate Generation & Delivery workflow (among others) on shows including the next installments of our amazing series like Money Heist, Selena and others. We’re excited to help even more productions this year, as we’re continuing to build support for more use cases and polish the experiences.

Let’s now zoom out and take a look at the foundation that supports the 20+ unique media workflows that we’ve enabled in the last two years, with more being added at an accelerating rate.

No single monolithic service would scale to support the various demands of this platform. Many teams at Netflix contribute to the success of Media Workflows Platform, by providing the foundations we rely on for many of the steps taken.

Media Workflows Platform (also known as “Content Hub”): a component that powers all of our media workflows. At a very high level it is composed of the Platform, UI and Partner APIs.

  • UI + GraphQL Services: facilitating various media workflows, one use case at a time, built with the help of the recently open sourced domain graph service implementation for the federated GraphQL environment.
  • Partner APIs: external partner APIs enabling integrations into Netflix media workflows.
  • Media Workflows Platform: a flexible platform that enables diverse, scalable, easy to customize production media workflows, built on the foundational tenets:

— Resource Management to associate files, assets and other workflows
Robust Execution Engine execution engine, powered by Conductor
State Machine defining user and system interaction
Reusable Steps enabling component reuse across different workflows

Media Inspection and Encoding: scalable media services that are able to handle various media types, including raw camera media. Use cases range from gathering metadata to transforming (change format) or trans-wrapping (trim media).

Universal Asset Management: all media with its metadata maintained in a common asset management system enabling a common framework for consuming media assets in a microservice environment.

Global Storage: global, fault tolerant storage solution that supports file-based workflows in the cloud.

Data Science Platform: all of the layers feed into the data science platform, enabling insights to help us iterate on improving our services using data-driven metrics.

Netflix Platform Tools: paved path services provided in building, deploying and orchestrating our services together.

We’ve helped productions manage and exchange many petabytes of media which is only accelerating with more usage of the platform. Some of our recent workflows in Editorial are in pilot on a handful of productions, our VFX workflows helped dozens of shows, Media Review assisted hundreds of shows and, our Archival workflows are used on all of our shows. While we’ve innovated on many workflows, we’re continuing to add support for more workflows and are refining existing ones to be more helpful. Our media workflows platform is a robust, scalable and easy to customize solution that helps us create great content!

Thanks for getting this far! If you are just learning about how production media management works, we hope this sparks an interest in our problem space. If you are designing tools to empower media workflows, we hope that by sharing our challenges and approaches to solving them, we can all learn from each other. Realizing common challenges inspires more openness and the standardization we crave. We’re really excited to see the proliferation of open APIs and industry standards for media transformation and interchange such as OpenTimelineIO, OpenColorIO, ACES and more.

If you’re passionate about building production media workflows, or any of the foundational services, we’re always looking for talented engineers. Please check out our job listings for Studio Engineering, Production Media Engineering, Product Management, Content Engineering, Data Science Engineering and many more.

Source link