1340372 - Support 3gpp:video-orientation in WebRTC

Status: NEW

(Core :: WebRTC: Networking, defect, P3)

Description

[Randell Jesup [:jesup] (needinfo me)]

+++ This bug was initially created as a clone of Bug #1300665 +++

Split off, since we landed support for abs-send-time and toffset in bug 1300665.

This would allow avoiding rotating video in some cases before sending or after receiving.  Mostly though not solely for Android/mobile (in theory a rotating monitor on desktop might use it, maybe).

Comment 1

[Bulk Bug Changes for mreavy's org]

Mass change P2->P3 to align with new Mozilla triage process.

Comment 2

[Guang-De Lin]

I left a comment on bug 984215 but seems like this is the bug to support video-orientation feature.

May I ask when this feature will be implemented or is this in the Firefox product roadmap? Our team is developing WebRTC mobile apps, but Firefox does not support video-orientation and display video without rotation. Without this feature, our apps must rotate video frames manually before sending them.

Comment 3

[christoph.eggert]

Just wanted to point out that the missing support of this flag may lead to significant performance issues. While the orientation is correctly negotiated in direct calls (could be an issue with SFUs to even orientate correctly) to be viewed normally, some mobile devices get severe framedrops from having to rotate the video material. So a device that is able to do full HD 30fps streaming when being connected with a Chrome browser might be on 13fps when connected with a Firefox browser. Especially in the wearable market this makes it a significant performance issue rather than a programming inconvenience.

Comment 5

[Dan Baker[:dbaker]]

Attachment: WIP: Bug 1340372 - Support Video Orientation RTP header and rotated frames.

Comment 6

[Jim Mathies [:jimm]]

We were unable to find any way to test this, as we could not identify a service that leverages it. Much lower priority (backlog) for now.

Comment 7

[Jon Ross :jonross@chromium]

Please see 1958429.

This is leveraged by iPadOS cameras using WebRTC. The physical cameras in current iPad Pros are inverted.

Reproduction:

• Start a Discord Video Call
• Have users on iPadOS as well as on Firefox on a desktop platform

For the Firefox user the iPadOS user appears upside down. For Chromium users the orientation is adjusted and the iPadOS user appears correctly

Comment 8

[Lennart Grahl [:lgrahl]]

libwebrtc built for iOS/Android makes heavy use of this. I'm not sure about Chromium on mobile - have you checked that? If Chromium on mobile does not use this, I might be able to provide an Android app.

Comment 9

[Shaofan Qi]

Zoom expects video rotation to be applied on the receiver side, the lack of this capability in Firefox is blocking our support.

Comment 10

[Jim Mathies [:jimm]]

Can someone provide a link to a conferencing service that leverages this? We could not find a service that uses it.

Comment 11

[Jim Mathies [:jimm]]

(In reply to Shaofan Qi from comment #9)

Zoom expects video rotation to be applied on the receiver side, the lack of this capability in Firefox is blocking our support.

Can you provide a link to test case we can use to reproduce the issue?

Comment 12

[Jim Mathies [:jimm]]

(In reply to Jon Ross from comment #7)

Please see 1958429.

This is leveraged by iPadOS cameras using WebRTC. The physical cameras in current iPad Pros are inverted.

Reproduction:

• Start a Discord Video Call
• Have users on iPadOS as well as on Firefox on a desktop platform

For the Firefox user the iPadOS user appears upside down. For Chromium users the orientation is adjusted and the iPadOS user appears correctly

This bug is not related to issues experienced on iPadOS.

Comment 13

[Chris H-C :chutten]

(whoops, already covered)

Comment 14

[Chris H-C :chutten]

(In reply to Jim Mathies [:jimm] from comment #12)

(In reply to Jon Ross from comment #7)

Please see 1958429.

This is leveraged by iPadOS cameras using WebRTC. The physical cameras in current iPad Pros are inverted.

Reproduction:

• Start a Discord Video Call
• Have users on iPadOS as well as on Firefox on a desktop platform

For the Firefox user the iPadOS user appears upside down. For Chromium users the orientation is adjusted and the iPadOS user appears correctly

This bug is not related to issues experienced on iPadOS.

If it's not, should we reopen bug 1958429?

Comment 15

[Jon Ross :jonross@chromium]

(In reply to Jim Mathies [:jimm] from comment #12)

(In reply to Jon Ross from comment #7)

Please see 1958429.

This is leveraged by iPadOS cameras using WebRTC. The physical cameras in current iPad Pros are inverted.

Reproduction:

• Start a Discord Video Call
• Have users on iPadOS as well as on Firefox on a desktop platform

For the Firefox user the iPadOS user appears upside down. For Chromium users the orientation is adjusted and the iPadOS user appears correctly

This bug is not related to issues experienced on iPadOS.

Per my previous comment, the Firefox user on Desktop is experiencing the bug of non-rotated video.

iPadOS was an example device sending the 3gpp:video-orientation data

Comment 16

[Shaofan Qi]

(In reply to Jim Mathies [:jimm] from comment #11)

Can you provide a link to test case we can use to reproduce the issue?

Do you intend to support this extension in sendrecv or recvonly direction?
I don’t have a test link available at the moment, but you should be able to inject or mock the header extension directly in WebRTC after negotiating it.

Comment 17

[Chris H-C :chutten]

Attachment: Screenshot 2025-12-11 at 20-56-21 Discord Echo Chamber Beer O'Clock.png

Here's a screenshot of a discord call where the top-right participant has connected using an iPad and the recipient (me) is using the Discord web app in Firefox 147.0b2 on Windows