EAI-708 support

[zsmatyas]

Previous version of the v1 and v2-dev branches had the entire SEI data put into a buffer and then read by the Eia608Parser. The parser had a check that the type and validity bits are set to channel 1 of the 608 content, and everything else was dropped. See: V1 if (ccType != 0) { seiBuffer.skipBits(16); continue; } Here the ccType values mean:

• 0 = channel 1 and 2 of EIA-608
• 1 = channel 3 and 4 of EIA-608
• 2 = Start of EIA-708 content (channel ID is defined inside the stream)
• 3 = Chunk of EIA-708 content

As only value 0 was handled, and no additional checks were used to separate channel 1 and 2: a single 608 channel was supported. I added my additional channel and EIA-708 support by updating the Eia608Parser to collect the data instead of dropping it, so channel 3-4 and EIA-708 bits were collected and parsed appropriately.

In the latest v2 code, all these additional data are dropped at SeiReader level: V2 like: if (ccValidityAndType != 0x04) { seiBuffer.skipBytes(2); } The change was: Link

SeiReader seems to be the correct location to filter 608/708 and channel information, but needs serious changes in the architecture I had. Are you guys planning to ever support EIA-708 (as it is an FCC requirement from 2017)? Did you consider supporting the other channels of EIA-608? Any similar changes like the one you made in the v2-dev branch requires lots of changes to support all subtitles correctly, that is why I am interested in the plans about supported formats.

Are you willing to accept pull requests related to these features or are you maybe already working on these issues?

Big Thanks!

[ojw28]

@cdrolle is looking at 708 support. Could you comment on the above? Thanks!

[zsmatyas]

Any news? :chicken:

[zsmatyas]

:honey_pot:

[zsmatyas]

Any comments about your implementation plans? I have the 708 parser implemented, I would gladly contribute if our code changes can be aligned. (As my V1 parsing is not compatible with the current state of V2)

[ojw28]

We have some 708 code internally that hasn't been pushed to GitHub yet. @cdrolle - Can you make it happen? Thanks.

[ojw28]

The change above adds what we've done so far, but it's not yet a complete solution. Some further work is needed on the MediaSource/Extraction side to expose the CEA-708 tracks so that they can be selected for rendering.

[roticv]

What's the current state of getting the github version up to date with the 708 changes?

[ebr11]

Hi. Has there been any further movement on this 708 support?

[roticv]

Last I have looked 708 decoder is found: https://github.com/google/ExoPlayer/blob/release-v2/library/core/src/main/java/com/google/android/exoplayer2/text/cea/Cea608Decoder.java

[zsmatyas]

@roticv That is for 608, not 708. I think there are a few pull requests pending for quite improved 608 and 708 support. Unfortunately, it seems they wont be merged in this form as reviewing them is quite difficult.

• 608 fixes and new features: https://github.com/google/ExoPlayer/pull/4900
• 708 fixes and new features: https://github.com/google/ExoPlayer/pull/4595

Reworking the pull requests into smaller digestible changes will take some time.

[sneelavara]

@ojw28 is there any plan to add full support for CEA-708. We have requirement to enable CEA-708. It would be helpful if this gets some priority.

@zsmatyas we have tried Sarnoff streams with #4595 changes. It works much better with that. But still there are many outstanding issues. We are happy to help in #4595 merge (and further improvement). Please let us know. Thanks.

[ojw28]

We don't have capacity to work on this any time soon. We would welcome high quality external contributions. In terms of what that means, I think we'd be looking for:

1. Small incremental changes that are relatively easy to reason about, and are amenable to careful code review
2. Test cases and/or content that (a) allow manual validation of the changes being made, and (b) can be used by tests

[sneelavara]

@ojw28 I have sent the test stream links over developer email. Please have a look at them. We are exploring the way to add unit test cases. If you have an inputs and/or pointers to look at, would be very helpful. Thanks.