FOMS 2018 — Standards, Unconference Style

The FOMS 2018 (Foundations of Open Media Standards and Software) conference took place mid-October here at Ellation HQ in San Francisco. FOMS included a key set of participants from the tech/publisher side and browser/platform side. Similar to last year, it was timed closely with Demuxed 2018 video engineering conference. Demuxed greatly increased in size this year attracting hundreds of video engineers from around the globe to discuss everything from per-title encoding, low latency streaming and adaptive video delivery with presentations from major companies in the space such as Netflix, YouTube, Twitch and many video tech startups.

This year’s FOMS covered a broad set of working sessions on standards that impact web video delivery. The unconference style is uniquely well structured towards addressing pain points that we, as video engineers, face in delivering against the growing expectations around quality of experience that our viewers demand.

White boarding MSE(Media Source Extension) concerns.

A standards effort is a usually complex undertaking as there is an ecosystem of competitive companies with entrenched interests that must be balanced and many unique use cases that don’t always fit with maintaining simplicity. A look at the standards process in W3C, for example, does not exactly inspire one to dive into this undertaking. FOMS serves as one of the contexts for more flexibility in early stages of these standards adjustments or new proposals. Some, like Siliva’s WebVTT proposal from years past, is now at W3C Candidate Recommendation stage and is widely used today. It gained much more broad adoption then more feature rich formats like TTML against its simplicity and similarity to community defined srt format.

Like in years past, there were many such early stage proposals and conversations such as John Bartos from JW, proposal for low latency annotations to the HLS spec that was well received relative to backwards compatibility with existing HLS streams, and its relatively straightforward approach towards ~2 second latency against a http protocol that normally has much higher latency. The unconference style enabled open discussion of the proposal with streaming software providers and other playback engineers. This enables companies like Apple or Google to share concerns they would have in supporting it within their browser or device ecosystem.

L-HLS (low latency hls) discussion at FOMS 2018 hosted at Ellation.

A big challenge we face at Crunchyroll is around fansub-defined standards for high fidelity anime subtitles not fitting into the industry standard format definitions that are natively supported across the devices and platforms we support. A common fansub format that we work with is advanced substation alpha.

Rafik Mikhael, a staff Engineer here at Ellation on the video infrastructure team, helped illustrate the challenge in a presentation around enhancements needed for real world WebVTT implementations to get closer to our subtitle content fidelity.

Anime film content is often inspired by or directly based off of manga comic books. Because of this translation from manga to film, we have found that a substantial set of anime content includes mixed media of Japanese characters as part of the cinematic language of the content. An artistic aesthetic is lost with captions formats designed for captions reliability vs high fidelity subtitles that enable the translator to preserve the original intent.

What Ellation renders today via application layer rendering or hard sub renditions depending on the capabilities of the device.
The WebVTT version as rendered by common WebVTT implementations.

Support for higher fidelity features in standard format such as VTT would enable Ellation to move away from costly per platform application layer renders or burned in subtitles that we leverage today, while maintaining the artistic intent afforded to our translators with high fidelity caption formats such as advanced substation alpha. Moving away from burned in subtitles has obvious benefits such as faster turnaround times on translations, low cost corrections and ability to support more languages as we expand our offering internationally.

We are proud to sponsor and participate in FOMS to further these and other goals. As you can see from the agenda this post only scratched the surface of what was covered. You can sign up for the FOMS mailing list or #foms channel in video-dev for updates on future workshops.

We hope to have a few follow up posts with more details around some of the other Ellation projects discussed at FOMS and other Engineering projects more generally; so be sure to follow Ellation tech blog for updates.

As always, Ellation is hiring so do shoot us a note if you would like to learn more about what we are up to and would welcome ideas about what’s next :)