You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There's no need to watch the entire clip looking for artifacts, just focus on the railing. In the first frame you can clearly see its grainy texture, then, despite not seemingly changing between frames, a part of the railing slowly loses details and turns into a blurry blob without any grain.
Here's the exact part which steadily loses details:
Video encoding is a terribly complicated algo with tons of parameters, so there's no rush in regard to dealing with this bug report. AFAIK a similar bug report for x265 was filed many years ago and was never addressed.
I have an encode at -qp 19 ready, I will check it later but it's substantially bigger (45 vs 60MB).
The text was updated successfully, but these errors were encountered:
So this is a somehow tricky question. You want to encode the keyframes with some more bits, because they will be used as references by all other pictures, and the quality of those keyframes propagates throughout all other frames. But of course you need to find the correct balance. If there is more random/uncorrelated movement/noise between the frames, encoding this better in an I-frame does not have the desired propagation effect. I think this is what is happening here. The encoder increases the quality in the keyframe, only to find out later, that the grainy stuff cannot be reused. We are looking into flattening the GOP structure for dynamic scenes and increasing the QP-differences for static scenes, but this is still work in progress or even planing stage.
x264/5 has a --tune grain parameter which lowers quantizer variation between frames but it also causes the bitrate to double. av1 has an interesting feature where they can filter out the grain/noise and render it back in at the player. I would assume this is like adding dither noise.
I've noticed that
I
frames have generally more details retained vs.B
frames which feature more blurriness.Is this intentional/by design?
Here's the resulting video.
Encoding settings used:
vvencapp --preset slower -qp 20 -i out.y4m --output=bit.266 # version 1.6.1
There's no need to watch the entire clip looking for artifacts, just focus on the railing. In the first frame you can clearly see its grainy texture, then, despite not seemingly changing between frames, a part of the railing slowly loses details and turns into a blurry blob without any grain.
Here's the exact part which steadily loses details:
Video encoding is a terribly complicated algo with tons of parameters, so there's no rush in regard to dealing with this bug report. AFAIK a similar bug report for x265 was filed many years ago and was never addressed.
I have an encode at
-qp 19
ready, I will check it later but it's substantially bigger (45 vs 60MB).The text was updated successfully, but these errors were encountered: