Manually generate "empty" h264 p-frame

Question

Let's call a p-frame frame empty if it doesn't change any pixels in the currently decoded video (i.e. no motion vectors, nothing).

What I need is to be able to manually insert empty p-frame into the video stream on request (need to send frames to the streaming client with a constant framerate, however the frame source on the streaming server can supply frames with a different/lower one).

So I need to be able to correctly form the byte sequence which represents the empty p-frame for the current resolution (I wonder which other parameters could be needed?)

Ideally, I prefer to have an encoder-independent solution, since I use two different ones: nvenc via Nvidia Video SDK and x264 via FFmpeg.

Where should I look to?

Saintpierre

This question is WAY to broad for stackoverflow. To point you in the correct direction, look into h264 bitstream parsing, VCL and CABAC coding and SKIP macroblocks. – Transfusion

@Transfusion okay, will look into FFmpeg's one! Should work identically for both encoders more or less, shouldn't it? – Saintpierre

also, just found this: h264bitstream – Saintpierre

Did you find a solution? – Microphysics

@DavidQHogan Sorry for late answer. I don't remember exactly, but probably just give an h264 encoder two (or more) indentical frames in a sequence and see its bytestream output – Saintpierre

Answer 1

I think there is a h264 nal unit dedicated to data padding (nal_unit_type: 12 : filler_data_rbsp( )). It might be useful for you.

Wintertide