BD Ripper (3D Plus) Tips to improve overall quality of conversion?

[bletour]

Buckle in.. this will get technical. I converted a blu-ray of Doctor Sleep and watched it on my big screen. I was disappointed with the overall quality. There were lot's of digital artifacts especially in low light scenes. That started me down the road of measuring the quality of various settings in version 11 and now version 12. Using the VMAF tool, I measure a few different clips of the video using various settings in dvdfab and then also against a FFMPEG encoded version utilizing NVENC since I have an Nvidia 20XX Turing card.

Here are some results:
LRLS = Lightning Record/Lightning Shrink (Nvidia 2070 Turing Card)
SW = Software Encoder (No acceleration)
SQ/HQ = Standard Quality/High Quality
1Pass/2Pass = Encoding Passes
VMAF is out of 100, PSNR out of 60, and MS_SSIM out of 1
FFMPEG used NVENC encoding/decoding with a 5Mb/s VBR and 8Mb/s Max
Clip 1

DVD Fab Version	Encoding Settings	VMAF Score	PSNR Score	MS_SSIM Score
11	LRLS HQ 2Pass	87.9577	44.5194	0.995780
11	LRLS SQ 1Pass	87.7419	44.5210	0.995719
12	SW SQ 1Pass	78.5874	42.3323	0.994097
12	LRLS SQ 1Pass	88.0855	44.3956	0.997606
FFMPEG	NVENC_8M_1Pass	94.6138	49.9948	0.998656

Clip 2

DVD Fab Version	Encoding Settings	VMAF Score	PSNR Score	MS_SSIM Score
11	LRLS HQ 2Pass	71.8197	30.5846	0.975837
11	LRLS SQ 1Pass	70.5031	30.5202	0.973363
12	SW SQ 1Pass	46.0894	27.0846	0.935390
12	LRLS SQ 1Pass	71.7805	30.5739	0.983692
FFMPEG	NVENC_8M_1Pass	95.7707	42.3496	0.995377

Clip 3

DVD Fab Version	Encoding Settings	VMAF Score	PSNR Score	MS_SSIM Score
11	LRLS HQ 2Pass	73.3437	37.1798	0.982553
11	LRLS SQ 1Pass	72.8649	37.1548	0.981832
12	SW SQ 1Pass	56.2030	34.3361	0.968179
12	LRLS SQ 1Pass	73.5675	34.3361	0.968179
FFMPEG	NVENC_8M_1Pass	94.4231	46.1366	0.997327

To ensure that the clipping wasn't a major cause (even though I was using a mostly frame-copy clipping tool) I also ran an analysis on the entire video with similar results:
Entire Movie

DVD Fab Version	Encoding Settings	VMAF Score
11	LRLS HQ 2Pass	78.5058
FFMPEG	NVENC_8M_1Pass	94.3652

I also repeated this test with two other movies and saw similar results. It is obvious to me that

1. SW Encoding is maximizing speed over quality. Not unexpected of since SW Encoding can take a long time on most HW. Would be nice though to have the different profile qualities be more inline with x264 in general. It is not very good at the default settings. CABAC is disabled, CBR 9Mb/s, 8x8DCT is disabled so on.
2. The LRLS settings for NVENC are not optimized for newer Nvidia HW. I also tried with a previous generation 1070 card with similar results
3. FFMPEG at only 5Mb/s is generating a much higher quality file than DVDFab at 9.1Mb/s when both using Nvidia for encoding
4. 2Pass Encoding isn't worth it and in general using HQ over SQ isn't worth it considering the change in file size
5. DVDFab 12 has cranked up the bitrate on SQ to 9Mb/s but I see very minimal difference in the overall quality improvement

I cannot find any way to update the settings used for LR or LS. Is there anyway to change those settings? I did find a setting for B-frame support for Turing cards and enabled it, but there was minimal difference.

Any hidden switches for the HW based encoding?

--Bill

[signals]

You can speak directly to the encoder in v12.x Ripper module, see this link: https://www.dvdfab.cn/tips/custom-h2...12_client_h264

You are the only user I remember that agrees with me that 2-pass is a waste of time. I have been preaching this since Fab v8.

[bletour]

Thanks signals. Although it appears these parameters only apply to x264 encoder (i.e. software). They don't apply when using Nvidia encoding do they? For instance, if I use ffmpeg with a raw blu-ray file, I can tell it to use Nvidia decoding/encoding and get FAR better performance than asking Fab to do Lightning Shrink.
For example this simple line:
ffmpeg.exe -vsync 0 -hwaccel cuvid -c:v h264_cuvid -i "<input video>.mkv" -c:a copy -c:v h264_nvenc -b:v 5M -max_muxing_queue_size 400 "<output video>.mkv

Takes a video and decode/encodes entirely with Nvidia with a max rate of 5M and the quality score is higher than using Fab. I can then tweak it even further to improve quality from there.

If I use the encoder directly, I think I am limited to SW encoding only correct? x264 doesn't have a way to tell it to use HW encoding that I know of. I was hoping there was a way to tweak the parameters of LR/LS.

I admit I only played with the parameter field a small amount, so maybe I'm just missing the obvious.
Thanks for any help!!!

--Bill