Best Rendering Practices/Settings Opinions?

[JDubyew]

Hey, guys. I finished my first approved fanedit from a HD Blu-Ray source using throwgnCpr's guide about a week ago (+10 cool points to those understanding the very subtle reference ) and I found the process to be pretty straight-forward to follow. But I wanted to get some second opinions on the optimal settings to use when rendering the lossless AVI and when using Handbrake.

a) When rendering the lossless AVI file in Sony Vegas using the Lagarith codec, what is the best color space to pick? I've heard that using YUY2 is much more space efficient than RGB, so I attempted to render in that first, but I deleted and re-rendered in RGB because I imagined video artifacts being produced due to not using RGB when it was just extremely minor film noise in the source itself heh.

b) When rendering a final .mp4 file in Handbrake, what do you guys think are the optimal settings? What does the x264 Tune, Profile and Level settings do (the Level setting in particular I'm interested in understanding)? What do you guys think is the best RF value? For my recent edit I had it set to RF 20 with a Profile of High and a Bassline of 4.0, no Tune set, and I was pretty happy with the rather lossless results leading to an 8GB 1080p file.

 

[hasmak]

Good article about H264 with information about profiles and levels on wikipedia
http://en.wikipedia.org/wiki/H.264/MPEG-4_AVC
I tend to use 2 pass average bit-rate with some tweaks to the default handbrake settings, this gives me better control of file size.
I generally encode 1080p with high profile @ level 4.1 and average bit-rate 3800kbps. this gives a file size of slightly less then 2GB per hour of movie with audio encoded at 320kbps.
I have some handbrake presets with the modified parameters, if you would like to have them send me a PM and I'll give you a download link and explanation of the modifications
Edit:
You might also want to have a look at:
https://trac.handbrake.fr/wiki/x264VideoSettings
and
https://trac.ffmpeg.org/wiki/Encode/H.264

 

[ssj]

excellent discussion, jdub & hasmak. hope to hear others weigh in on their preferences.

 

[JDubyew]

excellent discussion, jdub & hasmak. hope to hear others weigh in on their preferences.

Same here, it'll be interesting to see what people think!

Since I made this thread, I've had a look more closely at online comparisons of Handbrake settings, and have done more of my own tests. It seems that the "Level" setting is essentially a way for the program to limit the file size, and around 4.0/4.1 is the minimum required for maximum quality 1080p files. So it's essentially best to set the Level to anything above 4.0, but keeping it to 4.0 or 4.1 is best when rendering a file with the standard 24FPS framerate.

I also did a few renders using different Constant Quality settings, and found that setting it to RF15 is about equal to the original in quality and file size, whereas 16-20 maintains virtually the same visual quality whilst significantly downsizing the file size. Anything above RF20 tends to cause immediately noticable differences. For instance, a test render of my Redux fanedit at RF22 lead to dark scenes having macroblocking in areas of pure black. This is a nice comparison of RF settings I found.

As for two-pass versus constant quality, as far as I'm aware the former offers better file size control, but the latter is simply a better quality compression algorithm. Have a look at this blog post to see what I mean.

 

[hasmak]

Same here, it'll be interesting to see what people think!

Since I made this thread, I've had a look more closely at online comparisons of Handbrake settings, and have done more of my own tests. It seems that the "Level" setting is essentially a way for the program to limit the file size, and around 4.0/4.1 is the minimum required for maximum quality 1080p files. So it's essentially best to set the Level to anything above 4.0, but keeping it to 4.0 or 4.1 is best when rendering a file with the standard 24FPS framerate.

Actually, the profile and level setting is a way to tell the encoder which sub-set of parameters it can or can not use, this is particularly important for hardware players like WD TV Live, Seagate FAT, built in players in many new TV screens ..etc
As for Average bit-rate vs CRF, here is a fundamental difference in the algorithm; Average bit rate attempts to decrease the bit rate for frames that are not important, and increases the bit rate for the frames that are important; On the other hand, CRF attempts to distribute the information evenly across all frames.
Which one gives better quality? This a subjective matter, and we can argue about it forever, ultimately the descision to use one way or the other rests with you.
I believe that when encoding you must consider several factors:
First, you have to consider your source, how complex is it?, does it have fast moving scenes? does it have fast pan and zoom?
Does it have many dark scenes with details that need to be preserved?
These factors will decide what parameters other than bit rate you would have to optimize.
The second thing you will have to consider is your target audience, and your target viewing device.
I believe most fan edits will be watched by critical film enthusiasts, and will be viewed on some form of flat screen 40 - 50 inch using HDMI capable hardware, and I encode for that.
Having this in mind, I also believe that there is a point in the encoding parameters, beyond which adding more bit-rate will yield little or no noticeable difference to the audiance (low of diminishing returns); and in some cases may lead to negative returns when the file size becomes prohibitive to some people, or when the very high encode specs render the file unplayable on lower end or older hardware
I have uploaded a couple of presets to Google drive, you can have a look at hem here
Here is the link to the presets
https://drive.google.com/open?id=0B0...eEU&authuser=0
There is also a short clip that I encoded both ways as proof of concept, you can view the folder content and download from here
[link removed]

It would be nice if you would download and compare both clips and let me know the difference (I am probably biased, and my eyesight is not what it used to be:-|)
If you are going to test the presets, (and I hope you do) please note that the speed is set to medium, if you set it to slow, you will get a 10 - 15% improvement for the same file size
Also do comparison tests on small clips of your choosing with different inputs (fast motion, dark scenes ..etc)
If you do that, I'll setup a public folder where anyone can upload and download the test clips and give feed back.
After all, the proof of the pudding, is in the eating:lol:
Update:
I have uploaded 2 more encodes of the same clip using CRF @20 and @24.
Hope you will watch and comment

 

[hasmak]

I've added some more test clips to Google Drive.
I encoded the first 6 minutes of The Hobbit: BOTFA with average bit-rate ranging from 2800 to 9320kbps and constant quality CRF ranging from 28 to 20
I selected this clip because it has pan & zoom, fast motion, dark scene, high contrast colors, and rapid scene changes; that is to say, all the elements that would give rise to artifacts, problems or encoding defects.
It would be nice if you would watch the clips and comment on the quality, in particular, the lower bit-rate encodes.
To spice things up, I also uploaded the clip encoded with H265 using CRF@28. I'll also encode the same clip with VP9 and H265 using average bit-rate and upload it later

 

[addiesin]

I've added some more test clips to Google Drive.
I encoded the first 6 minutes of The Hobbit: BOTFA

I removed the link. You cannot link to a direct download of commercially available video. Feel free to give interested parties links in PM or (and I realize this may kill the point of what you're trying to do) upload to a streaming site like Vimeo or Youtube.

 

[hasmak]

I removed the link. You cannot link to a direct download of commercially available video. Feel free to give interested parties links in PM or (and I realize this may kill the point of what you're trying to do) upload to a streaming site like Vimeo or Youtube.

I understand your concern, however this was meant entirely as a tech demonstration of the results of different encoding parameters, and to encourage and stimulate a conversation about sensible encoding parameters for fanedits.
Uploading to Vimeo or YouTube would defeat the purpose as you wold not be able to download the clips in their original state.
Anyway, if anyone is interested, please PM and I'll give you the link.

 

[TV's Frink]

I understand your concern, however this was meant entirely as a tech demonstration of the results of different encoding parameters, and to encourage and stimulate a conversation about sensible encoding parameters for fanedits.

It's not his "concern," it's the rules, and he's enforcing them. The rules are in place to protect our site, and we enforce them in all situations. Please do not discuss it further, send a PM if you feel the need to discuss further.

 

[thecuddlyninja]

I watched all of the test clips and here are my thoughts:

Same clip, encoded different ways.
Constant Quality: RF 20, 22, 24 and 28
Variable Bitrate: 9320, 3800, 2800 and 1400 kbps

Constant Quality: RF 28 was quite bad, RF 24 was acceptable but started to show cracks in the darker spots. RF 22 looked good but the camera movements looked slightly stuttered to me. It was quite distracting. The definition was good, though. RF 20 looked really good but was slightly jittery as well. I could see everything in various lighting well. Worth the 25% file size increase from RF 22.

Average Bitrate: 9320 kbps looked great. 3800 kbps looked equally as good to me, and was under half the size of the 9320 kbps clip. I noticed no difference. With the 2800 kbps, I noticed a slight deterioration in definition of the darker things on screen, and darker scenes. The 1400 kbps test clip was noticeably worse. I could see some blocking and noticeable deterioration in definition during movement and darker lighting.

Overall Constant Quality vs Average Bitrate: I found the motion to be much smoother with the average bitrate clips. I'm not sure how to word this, but colors seemed to transition less with the average bitrate. What I mean is it seemed slightly more dynamic. You could argue whether or not faster color and lighting changes are more dynamic or more startling but I got the sense of the changes more with the average bitrate. There seemed to be no reason to encode at 9k kbps, the file size was twice as large as the 3800 kbps clip with no noticeable difference (to me, anyway). I suppose that's why I took hasmak's advice (and presets) and encoded both of my edits at H.264, variable framerate, avg bitrate of 3800 kbps. I think it's the best considering quality and filesize for dynamically lit movies with lots of motion.

 

[hasmak]

I watched all of the test clips and here are my thoughts:

Thanks for watching and for your feedback
I hope others will watch and give feedback, this could be helpful for many people