What the? Timeline Resolution impacts final video quality?

I just watched this video on YouTube:

Go to the 3 minute mark in the YouTube video.

Huh? Have all the videos I've been doing have degraded video quality? I always thought Timeline Resolution had NOTHING to do with the video quality you render. I always thought of it as sort of another version of Optimized Media, except that Optimized Media gives you files that are easier to playback, but at a reduced video quality. Similarly, I thought you could drop down Timeline Resolution, say, from 4K to 1080P, and your ability to traverse the timeline might be improved.

BUT at the end of the day neither Optimized Media nor the Timeline Resolution did anything to video quality when you went to go render. I could leave the Optimized Media as-is, leave the Timeline Resolution at 1080p, and then render at 4K. DR would then take the ORIGINAL 4K video and render at 4K directly without Optimized Media or Timeline Resolution impacting anything. Based on this YouTuber, I've been doing this wrong along. Is he right?

He is right. You can work in a lower resolution in the timeline all day long, but before rendering you'll need to change the timeline resolution. DR is even giving you warning about this, did you never read it?

Add Higher Resolution Renders?
The selected render resolution is higher than the
timeline resolution. Rendered images will be upscaled
from timeline resolution to the selected resolution.
This may result in reduced image quality.

You never saw this warning John? It's pretty clear.

I did, but I guess there are plenty of times that I am upscaling and either receive the same message or one that is similar and know that I should disregard it.

What's interesting is that I see others have fallen victim to the same thing I have and have a different understanding.



I think what also adds to the confusion is that when DR accepts media in a blank timeline, I many times get this window:



So DR seems smart enough to match the frame rate of the incoming video, but it leaves the Timeline Resolution as-is (i.e. 1080p if the incoming footage is 4K)? It seems like it would make more sense to match everything and then if someone had the knowledge to manipulate the timeline resolution to improve performance, then they would be the best candidates to know to increase it when going to render.

However, I do see your point about the window warning me of this when I do go to render, so its really no ones fault but my own. I guess the only good news is, is I'm comparing some of the footage where I had 4K footage and left the timeline resolution at 1080P and now comparing it changing the timeline resolution to 4K when I go to render. I'm really struggling to see a difference in quality. They seem identical.

Depends on your sources. Highly compressed UHD from a Bayer-sensor is not really 4K.
And then, what's your display? If it's an UHD TV, they have pretty good upscalers these days.

I can't test this as I'm waiting for a render to finish, but it's a simple test: Set Timeline resolution to 480*270 (1920*1080 / 4) render at 3840*2160. If the image looks crappy, timeline resolution has an effect. if not, it does not


Edit 2: Just tested. 1) resolve says as follows (see image)
2) The resulting image is scaled from timeline resolution to output resolution.

See Resolve 18 manual chapter 184 Delivery Effects Processing, page 3864.

I guess the only good news is, is I'm comparing some of the footage where I had 4K footage and left the timeline resolution at 1080P and now comparing it changing the timeline resolution to 4K when I go to render. I'm really struggling to see a difference in quality. They seem identical.

That could be explained by how you're viewing or the nature of the footage -- prevailing viewing conditions make 4K delivery ludicrous most of the time -- but if you want to see this principle in action, set the timeline to a *really* low resolution, like SD, and then export to 4K.

What's happening is, the original 4K files are downsampled to the timeline resolution, and then the Deliver page upsamples the files at the current timeline resolution to delivery resolution. This is not good.

[EDIT: I see Mads beat me to it. If you perform his experiment, you'll see it loud and clear.]

Everything is rendered at the timeline resolution. So if your export is different, scaling will have to occur.

Thanks for the replies. Funny, I did sort of what others said before reading these comments and it did prove what was said, that is, the timeline resolution definitely impacts video quality if they don't match up before rendering.

As a test I took a small piece of 4K video and rendered at 4K with a Timeline Resolution of 720P and then with a timeline resolution at 4K. Here are the two videos for comparison.

720p Timeline Resolution:
https://drive.google.com/uc?export=download&id=15Kg1kJBLZULVQX4QvDkIPtY7oMcj424P

4K Timeline Resolution:
https://drive.google.com/uc?export=download&id=15LlONMDYUtML3m69H207rI3_WbqtLF0h

I was surprised to see how close they were and wondered if my eyes were playing tricks on me, so then I blew up a still image of the video at the same point in the video. It then became a little more clear that one of the videos was softer than the other. It was also sort of obvious before all this because of the difference in file size. The 720p timeline resolution video rendering at 4K is 125 MB whereas the one with a 4K timeline resolution and still rendered at 4K is 250 MB. There is clearly more information there. Still, I'm a little surprised at how slight it is, at least to my eyes just looking at the raw video.

720p Timeline Resolution:


4K Timeline Resolution:

Everything is rendered at the timeline resolution. So if your export is different, scaling will have to occur.

That's a good point, but its interesting to me that the render times are identical. I would have expected if scaling was introduced then that would push out the render time further, but that isn't what happened in reality. The two little videos above (one with a timeline resolution of 720p and the other at 4K, both using raw 4K footage and rendering at 4k) rendered at one minute and 51 seconds exactly, both of them. Again, sort of weird that the addition of scaling didn't seem to impact render time, but it certainly did file size.

You haven't supplied any information about how you're exporting the material, but if it's to a h.264/5 format in Resolve, the test is pointless. You're not going to see how stark the differences are in those formats in Resolve, which doesn't perform great long GOP compression. And you haven't indicated what data rates you chose for export.

I didn't download your files, but the screenshots above are about the worst test of resolution you could choose, because there's so little of it. You never going to see the difference with that kind of material.

You haven't supplied any information about how you're exporting the material, but if it's to a h.264/5 format in Resolve, the test is pointless. You're not going to see how stark the differences are in those formats in Resolve, which doesn't perform great long GOP compression. And you haven't indicated what data rates you chose for export.

I didn't download your files, but the screenshots above are about the worst test of resolution you could choose, because there's so little of it. You never going to see the difference with that kind of material.

Well, based on what others were saying, the difference would be dramatic and there was no mention of needing to choose a specific codec. I'll have to look up "great long GOP compression" because I'm not entirely sure what that is.

But to answer your questions, I did use H.264 and BEST for data rate in both cases. It seemed more important to try and keep things constant between the two examples so that ONLY Timeline Resolution would be the variable.

The screenshots were intentionally small so that you can magnify them to see the difference. I'm not sure what the point would have been to choose more of the image. You would still have to zoom in on the same area to see the difference. I do see though that it requires anyone reading this to make the images bigger. I'll make them bigger and then reattach them if that helps. That might allow others to see the difference.

To expand on this topic. I purposefully kept my timeline at 1080p and delivery tab at 4k for years when I had my Gtx1650 GPU.

With only 4gb of Vram, it was crash city if I tried to do a 4k timeline. Upscaling in the deliver tab seems essentially free, and you need 4k to upload to YT. So easier to do it in Resolve than in Handbrake.


Now I have a 16gb 3080 laptop so I can do 4k. I have two presets made, one at 1080p that is default for new projects. It has 'nearest' for retime mode and all the other fast settings you could choose from the project settings.

I then switch to a 4k preset when i go to export. Or in reality, I forget to change and when I get the popup on deliver tab of mismatch resolution, then I go change it.

I would have expected if scaling was introduced then that would push out the render time further

Not sure scaling at that point is very high quality. Compared to something like SuperScale, anyway.

Hardware may also be a factor here. Could be yours is sufficient to cover the additional processing.

Final scaling is using regular resampling and that is computationally a total non-issue for modern hardware.

I always thought Timeline Resolution had NOTHING to do with the video quality you render.

No, the opposite. Time resolution has EVERYTHING to do with the video quality you render. All this is explained in Chapter 11 of the Resolve 18 manual, "Image Sizing and Resolution Independence," starting on p. 259. That's a very important chapter and I tell students to read it twice so that they'll understand what's going on. Definitely some mission-critical information there.