Davinci encoded output is distorted and blurred

Disclaimer: sorry that I didn't make links out of PNGs but forums is blocking those for me Please do copy them and open up in your browser

Hi everyone!

So essentially I face this problem where most of my output is slightly (but noticeably!) lower in quality than my original recording even though I use same bitrate for both. Couple of screenshots:

Here's the original 1440p 25k bitrate encoded footage frame:
calculationwiki.waw.pl/fh4/org25k.png

Here's what Davinci gives me when I use settings presented below (H.264):
calculationwiki.waw.pl/fh4/davinci-25k-264.png

Here's what it gives me when I use settings presented below but with H.265:
calculationwiki.waw.pl/fh4/davinci-25k-265.png

Last but not least - my settings:


So if you open up the screenshots you can see noticeable drop in quality (fence in the right is losing it's structure way earlier, trunk of the tree on the hill on the left is getting blurred with the leaves on the ground, pavement texture is blurred etc). H264 is a tad bit better but H265 is losing even more quality.

Most of the time my editing is simple - remove non-interesting parts, add my watermark, sometimes camera, change volumes, that's it, no fancy color bumping/filtering or anything. However I do face this issue even though I do have the same bitrate output as original file (of course for 1440p we can get even higher, but there's youtube compression which will mess it up a lot anyway, so I stick with 25k).

Anything I am doing wrong here? Cause there must be something for sure.

Thanks in advance!

Change from Restrict to Automatic>Best.

Why is entropy set to CAVCL instead of CABAC? Also, you are encoding to high resolution @ 60 fps, you simply need more bitrate.

Why is entropy set to CAVCL instead of CABAC? Also, you are encoding to high resolution @ 60 fps, you simply need more bitrate.

Thanks a lot! But why do I need more bitrate if I want to keep quality the same as source material?

Thanks in advance!

Why is entropy set to CAVCL instead of CABAC? Also, you are encoding to high resolution @ 60 fps, you simply need more bitrate.

Thanks a lot! But why do I need more bitrate if I want to keep quality the same as source material?

Thanks in advance!

You can't keep the quality of the original because it's already (rather heavily) compressed. It's called generational loss. You need substantially higher bitrate than the original to keep the visual likeness.

You may want to use a better encoder too, like x.264.
Export a huge DNxHR file first and then compress in Handbrake. Slow, but good.

You can't keep the quality of the original because it's already (rather heavily) compressed. It's called generational loss. You need substantially higher bitrate than the original to keep the visual likeness.

I am rather familiar with such process TBH, just didn't know this is THAT noticeable when you keep the compression low.

So how much is "substantially" here? 120% of original? 200%? And what would be the best course of action if I want to have video on as-close-as-possible to let's say 25k bitrate?
a) Record in 40k then store as 25k (probably not, however there's more chance for better compression in 25k results if original material was better)
b) Record in 20k then store as 25k (it should give some headroom to keep the quality but if original is more compressed then offspring will also look worse?)

As said, I'd like to keep output rather small (25k) but pack as much of the fine detail there as possible.

[...] And what would be the best course of action if I want to have video on as-close-as-possible to let's say 25k bitrate?
a) Record in 40k then store as 25k (probably not, however there's more chance for better compression in 25k results if original material was better)
b) Record in 20k then store as 25k (it should give some headroom to keep the quality but if original is more compressed then offspring will also look worse?)

I've done a simulation. 40k "raw" does produce significantly better image when compressed to 25k, guess I'll stick to it. Now I just have to find sweet spot with how much of raw I want to have to achieve best results when going to 25k in Davinci.

H264 is just a tad bit better than H265 in some cases H265 in other, but if someone has any source to back me up here, then please do.

You may want to use a better encoder too, like x.264.

I've always used Q16 in x264 for my deliverables. I find it to be excellent quality.

I recently discovered that Resolve's "Best" is slightly better than that, so...

I want to have video on as-close-as-possible to let's say 25k bitrate?

Why?

I want to have video on as-close-as-possible to let's say 25k bitrate?

Why?

Cause I am producing a whole lot of videos that are 1440p and upload them mainly to youtube. Youtube does cut quality by a whole lot and adding about 100+ GB each month of videos to my archive does stress my archives a lot already. Disk space is not a big issue now but it would be if I'd up my quality to let's say 50-70k with not-so-much quality added that I don't really need

I'd say it's a "golden spot" for most part for Youtube - that's it.

I've always used Q16 in x264 for my deliverables. I find it to be excellent quality.

I recently discovered that Resolve's "Best" is slightly better than that, so...

Yet it may be waaaaay heavier Nice tip though

It's hard to maintain quality and avoid storage issues. You are practically going through a triple generational loss... Capture - export - youtube.

Sent from my Mi 9T using Tapatalk

Yet it may be waaaaay heavier

In my tests, it wasn't. And because the bitrate isn't predefined, you won't be getting more bits than you need to achieve the Best Quality.

Size depends a lot on content here. High motion combined with high detail over large areas will need more bitrate or you won't like the quality. You can't eat the cake and keep it.

In my tests, it wasn't. And because the bitrate isn't predefined, you won't be getting more bits than you need to achieve the Best Quality.

It all depends on the game, but if I stick to Forza it probably won't get that low since screen is changing all the time there. Especially since I've cranked my recordings to 40k now and only then I am doing the heavy encoding to 25k in Davinci (which will be most "harmful"). Unless CQP file gets less re-encoding (which too is very doubtful if you do ANY editing) then after deleting RAWs you'll be in the same place

Why Q16 tho?

Size depends a lot on content here. High motion combined with high detail over large areas will need more bitrate or you won't like the quality. You can't eat the cake and keep it.

Very true, but of course you have to know what dimnishing returns are and stop at ~85-90% quality cause afterwards storage cost is too high.

It's hard to maintain quality and avoid storage issues. You are practically going through a triple generational loss... Capture - export - youtube.

That is somewhat unavoidable if you edit the footage at least slightly. What *may* look better (depending of how VP9 will treat you) is to render twice - once for best quality and upload it to youtube, then smaller CBR to archieve it. But then this gets more time-consuming