Your Favorite "Set and Forget" Handbrake Settings?

[bmichaels556]

The top part of this is sort of leading with my conclusions. Below the dotted line is me rambling about my results with assorted videos, and why I chose them. Looking to hear your guys' perspectives!!

 

My ultimate standard is this: If file size is reduced significantly, and maintains quality to the point where a professional who deals with video every single day would have to sit down, take screenshots and stare at their screen from a couple inches away to really tell the difference, that's a win in my book. I'm probably at the high end of beginner or low end of intermediate and it seems like RF 20 with slow has been very nice for most, which is why I figure going down to 19 will improve a bit more and be "good enough" for almost anything. It ends up with about 1/3 to 1/4 the file size, and that's nice. RF 20 is a bit less overall, but I feel like I can tell a difference if only slightly. But that could just be placebo. Looking at them blind? I highly doubt I'd know the difference. Other misc settings... I go with 128kbps AAC for audio... And always sticking with MP4 with "web optimized" and "align A/V Start" always checked. And of course always leaving decomb and all that nonsense off unless needed.

 

For those of you who have used handbrake extensively, what has achieved the aforementioned standard in your opinion? I know it's subjective, but I think my expectations that I just mentioned are least objective enough as to be understandable and not totally out of left field... But this doesn't even go into using something like Sony Vegas to compress video, which I've found to have similar quality at similar bitrate, but have much lower render times. Half, even. So what the hell is up with that? I thought I knew, but I just don't!  

 

--------------------------------------------------------------------------------------------------------------------------------------------

 

So personally, I'm always taking video, both in real life and when gaming. And that means a ton of storage space needed. It's not professional or anything, but it's just something I like to do. And so I find myself using Handbrake quite often to get my file sizes to be reasonable. For you guys, what have been your best "set and forget" settings for different resolutions for decreased file size with basically no discernible loss in quality?

 

For all video, I've went with encoder preset "slow", no fast decode, and "auto" on profile and level. My reasoning for "slow" is that it doesn't REALLY add a ton more time to renders over Medium, but I notice almost no real difference between "slow" and "placebo". Originally I thought this was only about efficiency of file size, but I was dead wrong. It actually mattered for quality, and it surprised me. But slow seems comfortable and fast enough for my uses. I'm also doing it all h.264. My reasoning? While h.265 leads to the same quality at lower bitrate, or higher quality at the same bitrate, it also takes twice as long to process, and the difference in file size in my experience, is nowhere near half, as it should be judging by the time of completing the task. A 200 MB file in h.264 may end up at like ~150 MB for same quality but double the time to complete. I'm not so sure that's worth it with my current hardware (dual Xeon X5675 rig, 12 cores / 24 threads).

 

For 1080p60 gameplay recordings (I pretty much record all my PC gameplay), I'm starting off with 50 mbps HEVC on Radeon ReLive. With that setting, I've noticed very little difference compared to higher bitrates. I probably could start lower and still be just fine, but I figure if my source file is higher quality, the end file even after compression will be better, even at the same quality settings. At least that's the theory. Even still, if it's all getting compressed anyway, it doesn't make a difference if the initial file is larger, at least in my opinion. I believe Handbrake recommends RF 20-23 for HD sources, but I've found RF 20 to be noticeably better than 23, at least to my eyes, all other settings being equal. At RF 20, auto settings, and preset to "slow", I'm doing about 1/3 the file size, often less, as good as 1/4 the size.

 

It gets a little more interesting with older video from my point and shoot, especially the stuff shot at 720p, though even the 1080p stuff is pretty "soft". But RF 20 with all previous settings being the same, I do maybe notice just a slight difference between source and the handbrake compression. Not a ton, but something. Text maybe looks just ever so slightly worse. The biggest thing is that the colors do change a bit, and so that's another thing that distracts and makes it hard to tell what's changed and what hasn't when it comes to quality.

 

It seems like the higher the bitrate and resolution of the source, the better the final result will look at equivalent setting. For example, it does appear that 720p looks worse compared to other 720p videos, even when comparing 1080p to source at a given RF.

 

I originally started messing with just setting a constant bitrate, but like many of you will probably agree, that constant bitrate is more than needed for some scenes, and not nearly enough for other scenes. And that doesn't even go into differences between types of video (for example, high framerate gameplay recordings, live action videos/movies, animation and so on). I'm really thinking of just leaning towards leaving "slow", with profile and level always at "auto", and maybe an RF of 19 for everything. It seems like regardless of media type, if source resolution and bitrate are in the same ballpark, the decrease in file size is also fairly similar in my experience.

 

 

 

[LogicalDrm]

I usually just use default for 1080p YouTube preset. Though filming stuff at 1080p60 means I probably should make my own preset to shrink files before uploading.

[bmichaels556]

14 minutes ago, LogicalDrm said:

I usually just use default for 1080p YouTube preset. Though filming stuff at 1080p60 means I probably should make my own preset to shrink files before uploading.

Yeah the YouTube 1080p60 preset is definitely good, although I'm not sure why it's coming with decomb as default... At least on my current version. 

 

Depends on internet speed though. I'd get the highest bitrate file I can possibly tolerate to upload because it's going to get killed with compression anyway. So I do my best to keep my source as good as possible, I think there's enough of a difference. Or at least I'd like to think so.  

[LogicalDrm]

Just now, bmichaels556 said:

Yeah the YouTube 1080p60 preset is definitely good, although I'm not sure why it's coming with decomb as default... At least on my current version. 

 

Depends on internet speed though. I'd get the highest bitrate file I can possibly tolerate to upload because it's going to get killed with compression anyway. So I do my best to keep my source as good as possible, I think there's enough of a difference. Or at least I'd like to think so.  

Yeah, well. My current upload is only 5mb, so 17gb file directly from Resolve is huge nope.

[WereCat]

Yesterday I tried reencoding one movie in Handbrake. 

Source was h.264 BlueRay 38GB.

 

My settings :

 

Source resolution 

Align A/V at the start 

MKV

x264 (CPU) 

Medium Preset

Main Profile

CRF 19

 

Result 4.4GB

I can't really tell the difference vs source even after multiple screenshot and compares. 

 

Took about 55min to encode with 3900X. The video is 2h 20min long. 

 

While I'm happy with quality, I want to reduce the file size still... so I want to experiment with AV1 but encoding that will likely take days. 

[19_blackie_73]

1 hour ago, WereCat said:

Yesterday I tried reencoding one movie in Handbrake. 

Source was h.264 BlueRay 38GB.

 

My settings :

 

Source resolution 

Align A/V at the start 

MKV

x264 (CPU) 

Medium Preset

Main Profile

CRF 19

 

Result 4.4GB

I can't really tell the difference vs source even after multiple screenshot and compares. 

 

Took about 55min to encode with 3900X. The video is 2h 20min long. 

 

While I'm happy with quality, I want to reduce the file size still... so I want to experiment with AV1 but encoding that will likely take days. 

Get a small part of the movie with avidemux (at least I think I remember that avidemux can split without reencoding) and try with the small part. Way faster for testing

[WereCat]

2 hours ago, 19_blackie_73 said:

Get a small part of the movie with avidemux (at least I think I remember that avidemux can split without reencoding) and try with the small part. Way faster for testing

I just use ffmpeg to split video. 

 

I took a 10s part from a 120FPS 1080p game capture and it took over 4h to encode it to AV1 with the ffmpeg with the best quality preset. 

[bmichaels556]

13 hours ago, WereCat said:

Yesterday I tried reencoding one movie in Handbrake. 

Source was h.264 BlueRay 38GB.

 

My settings :

 

Source resolution 

Align A/V at the start 

MKV

x264 (CPU) 

Medium Preset

Main Profile

CRF 19

 

Result 4.4GB

I can't really tell the difference vs source even after multiple screenshot and compares. 

 

Took about 55min to encode with 3900X. The video is 2h 20min long. 

 

While I'm happy with quality, I want to reduce the file size still... so I want to experiment with AV1 but encoding that will likely take days. 

Holy crap, AV1 takes THAT LONG!? To be honest, I wouldn't care how much "better" it is, that demand for time is just way out of proportion for any benefit it might bring. Just my opinion though. Might I feel different if I had a quad Xeon rig and were doing these things professionally? Maybe. But even still, that's a lot to ask.. Not bad on those 3900X encode times though.

 

I ran some further testing for my own curiosity. I used some video from my Galaxy Note 4, which at 1080p seems to only record at about 15mbps, which is fairly reasonable for the quality you get from it in my opinion. Using all the settings I'd busted my ass to lock down, at RF 19 it ended up at like 13.5mbps!! So all that tells me is that it radically differs depending on quality of source file. Handbrake ultimately seems to ask, "What bitrate do I need for this level of quality? I don't care what the source bitrate is, I'm worried about source output for a certain level of quality." Hence the results I guess. Very interesting.

 

Further, cranking all the way down to RF 24 still shows relatively little difference from source but with much more reasonable bitrate. Compared to even Google Photos compression, equivalent constant bitrate ALSO is MUUUUCH better. Tons of nuance to this. It's been a fun adventure, but also frustrating. 

[WereCat]

6 hours ago, bmichaels556 said:

Holy crap, AV1 takes THAT LONG!? To be honest, I wouldn't care how much "better" it is, that demand for time is just way out of proportion for any benefit it might bring. Just my opinion though. Might I feel different if I had a quad Xeon rig and were doing these things professionally? Maybe. But even still, that's a lot to ask.. Not bad on those 3900X encode times though.

 

I ran some further testing for my own curiosity. I used some video from my Galaxy Note 4, which at 1080p seems to only record at about 15mbps, which is fairly reasonable for the quality you get from it in my opinion. Using all the settings I'd busted my ass to lock down, at RF 19 it ended up at like 13.5mbps!! So all that tells me is that it radically differs depending on quality of source file. Handbrake ultimately seems to ask, "What bitrate do I need for this level of quality? I don't care what the source bitrate is, I'm worried about source output for a certain level of quality." Hence the results I guess. Very interesting.

 

Further, cranking all the way down to RF 24 still shows relatively little difference from source but with much more reasonable bitrate. Compared to even Google Photos compression, equivalent constant bitrate ALSO is MUUUUCH better. Tons of nuance to this. It's been a fun adventure, but also frustrating. 

I did some more testing yesterday.

 

So my settings I mentioned above the x265, Medium CRF 19, 1-pass results in a 4.65GB file in about 55min encode time.

If I go with a 2-pass x265, Medium, 3000avg bitrate I get a 4.07GB file in about 2h encode time.

With a GPU H.265 encode on 1080ti with 3000avg bitrate and Highest Quality preset I get a 4.03GB file in about 13min encode time (which looks noticeably worse than the other ones).

 

Now I still want to try using slower quality preset and higher CRF which should give me even lower file size at simmilar quality but with a lot longer encoding times.

 

 

Regarding the AV1.

The libaom-av1 in the FFMPEG is that slow because it barely uses the CPU but its of a very high quality.

 

When I use SVT-AV1 I can get a 100% CPU usage and depending on the settings I use I can get anywhere from 50x to 3,000x faster encoding time vs the libaom-av1.

 

However I used only some SVT-AV1 GUI from github which has limited settings and I cant compare it directly with the libaom-av1 as it wont allow me to encode at specific bitrates.

I can actualy do SVT-AV1 trough FFMPEG the same way as I do the libaom-AV1 in FFMPEG but I need to figure out the commands yet. There actualy is a bitrate setting.

 

 

Just as an example, this is a 1080p 120FPS libaom-AV1 short clip I encoded from a gameplay. It is 2,000bitrate and of much higher quality than 1080p 60FPS at 10,000bitrate with x264. The source file is 150MB x264 NVENC at 100,000 bitrate.

 

 

AV1.mkv

 

 

Regarding the source file.

Yes, better source will give you better quality on the output with the same settings.

[Glassed Silver]

22 hours ago, bmichaels556 said:

Yeah the YouTube 1080p60 preset is definitely good, although I'm not sure why it's coming with decomb as default... At least on my current version. 

 

Depends on internet speed though. I'd get the highest bitrate file I can possibly tolerate to upload because it's going to get killed with compression anyway. So I do my best to keep my source as good as possible, I think there's enough of a difference. Or at least I'd like to think so.  

Better input is DEFINITELY a good way to ensure that YT's recompression causes the least possible harm. (although it will of course greatly affect you, no question)

 

Relevant MKBHD video