NvEnc Dirty Rectangles

DXGI DDA supports Dirty Recs and dirty moves. I want to implement this within my NvEnc implementation for better encode times and less data transfer. Is this possible with NvEnc? Is there a way to directly inform NvEnc of the dirty recs? Or do i have to manually implement this?

Also, is this even a reasonable feature enhancement? Would the additional overhead cause more latency?

‘NV_ENC_PREPROCESS_FRAME’ is mentioned in nvEncodeAPI.h and seems related as well, but the only place its talked about is in NVENC_RECT structs defintion, which is never mentioned again as well.

encoder only considers changes when encoding. you dont have to provide them and neither is there a way to do so.

Closest feature to this emphasis map. Emphasis map hints the encoder as to what parts of the frame are more important vs less important. encoder considers this hint when allocating rate control budget. for perfect images like desktop windows, it almost does nothing.

For example, if you are streaming on twitch and you have a camera on the bottom left corner. you can set emphasis map to give more priority to that space so that, in low bitrate conditions, the game might look low quality but your camera will always look best.

Well, this is also that whole using B frames as reference after ffmpeg fixed it after 2 years. It makes picture much better.