Darker optimized media

Hello, i'm trying R12B2 with a little project.

I'm adding GoPro Protune MP4 files to the media pool and made Resolve produce Optimized Media. If i switch the option "Use Optimized media if available" ON/OFF, i can notice that the original footage and the optimized media don't match on luma levels.
Where i'm wrong?

Thanks! Umberto

Optimised media is the version with lower bitrate, meaning there is much less information about colours as well as luminance. The picture is flattened and artifacted. If you posted still of original vs optimised along with scopes screenshots it would say more about your issue.

Of course yes.

Attached two screenshots comparison. On the left is the original media, on the right the optimized one.
I choosed an high end encoding for the optimized media (i guess a 444 encoding can replicate a GoPro footage color pretty well) but i scaled down to 1/4 due to performance issues.

Im perfectly fine in grading with a lower res footage, but if i use a darker one the result when rendered (which is based on the original media) will be washed out.
Shouldn't i grade on optimized media?

Thank you! Umberto

Woops, i forgot the scopes.
I'll post a more useful screenshot asap.

Maybe i should post a screenshot without grading, but in the previous it's still visible a difference. The grading with optimized media has more vibrant colors and a little deeper shadows.
A little more information: the GoPro footage is UHD 4:2:0 at 65mbps.

At first, i transcoded it to Cineform format to being able to edit smoothly in Resove. While using UHD Cineform footage is ok on my machine, it starts to kill my machine especially when i put some effects like lens distortion correction (obviously). Then i choosed to activate proxy rendering but it made the workflow really slow because the background proxy rendering is taxing my machine so much. Then i activated optimized media, and editing went smoother again (because of the reduced resolution).

But at this point, i decided to import in the media bin the original GoPro footage directly, without Cineform encoding which becomed an useless transcoding layer using optimized media. Then now the optimized media is computed on the original GoPro footage, which is converted from 4:2:0 65mbps to DNxHR 444 which, i guess, should be enough.

One more thing, i discovered that i needed to turn off automatic value range detection on each clip, i seen that by setting 0..1000 what is displayed on the video monitor matched with the final MOV H.264 rendering. But by activating the optimized medie the video monitor shows a slightly different image which doesn't match with the final rendering.

Umberto

Here i am with a more useful (i hope) screenshot.

Grading bypassed, in the upper end with original footage (and scopes), in the lower end with the optimized media (and scopes).

Loos like Resolve computes the optimized media by stretching the levels a little bit.
What can be the culprit?

Thanks! Umberto

It looks like the original media is video level and the optimize Is full data level. (Or viceversa).

Can you force the gopro to video level/data level and see if switch the contrast?

It looks like the original media is video level and the optimize Is full data level. (Or viceversa).

that's obvious...

but the more important question is, why it gets interpreted in a different way by resolves main processing and by the optimization transcoding subprocess?

Hello Walter,

All clips were initially in autodetect mode (and the result was worse). Then i switched them all to data level, and the produced optimized media is as shown (better). I just tried to switch the screenshot clip to video level and the optimized media is even darker. I guess that the autodetect goes for video level, as the result looks similarly bad.

Is the culprit the GoPro MP4 encoding?

All clips were recorded in ProTune/flat mode, which should output something similar to a logaritmic luma. I expect to see the very same washed out image in Resolve then grade it. I don't think that the flat mode would encode an "out of data levels" image.

Looks like it's necessary to apply a LUT when the optimized media is rendered?

Currently, i added a timeline grading that barely grades the original media similarly to the optimized media. Then i edit with the timeline grading off and optimized media on, and when i need to render i turn off optimized media and turn on the timeline grading, so i get a final render similar to what i see while grading. But it's a little bit cumbersome workflow.

Puzzling... Umberto

It looks like the original media is video level and the optimize Is full data level. (Or viceversa).

that's obvious...

but the more important question is, why it gets interpreted in a different way by resolves main processing and by the optimization transcoding subprocess?

I dont know... It was just a check....