ExternalEncoder for H.264 only used when video > 500x500x, otherwise OpenH264 encoding kicks in?

[Dominik Balogh]

In WebRTC internals, I can see that the External Encoder (hardware accelerated) is only enabled when the video fed to WebRTC is above about 510x510px. When the video origin is smaller than that, the OpenH264 encoder kicks in automatically.

Why is this?

My theories:

- It makes less sense on smaller video due to CPU overhead

- Maybe a Mac OS/drivers/GPU related issue

- GPU is unable to perform encoding on small video due to its H264 encoding/profiles implementation

- anything else?

Can this be fixed?

Tested on a variety of MacBooks/Mac OS/latest Chrome.

Thanks

Dominik

[Bo Zhou]

Hi

Can you try to pad the captured image to keep the resolution larger than the restriction and send it to the encoder ?

Thanks, I'm also newbie on H264 of WebRTC.

[Dominik Balogh]

What do you mean by "pad"? Can you share your idea in more detail?

Thanks

[Bo Zhou]

Hi

Pad means, once the captured frame is less than the constrained size, the video encoder creates a large frame buffer and send it to the encoder.

--

---
You received this message because you are subscribed to the Google Groups "discuss-webrtc" group.
To unsubscribe from this group and stop receiving emails from it, send an email to discuss-webrt...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/92916823-2628-403e-823b-8679422d668d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Dominik Balogh]

Interesting idea! So how would I do that? Thanks!

To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/CAMX42NBuGAk57fUy_5Cw6R4f7Uv-bWSE5Xf0Lcx5wuZSuwoT7w%40mail.gmail.com.

[Bo Zhou]

Hi !

You can use libyuv supplied by WebRTC

Here are document

https://chromium.googlesource.com/libyuv/libyuv/+/HEAD/docs

Let's use fixed size, such as 720P, but the incoming size of captured video frame is smaller than this size, we could copy the captured video frame to the preserved large size frame buffer, then give the frame to encoder.

Just an idea, hopefully it works for your case to prevent crash and cover special cases.

To view this discussion on the web visit https://groups.google.com/d/msgid/discuss-webrtc/jy9iod3k.f4b8f708-3c03-4328-8cc8-c7df371a5c56%40we.are.superhuman.com.