Blackmagic Forum

How do I get 4K movies to be small in size and still sharp?

I have compressed 2 hour 4K movies (Hollywood/real movies) that are 5GB and sharp as heck.

I shoot a 15 minute movie on my Sony Pro video camera and to render it sharp and good looking its 3-5 GBs. This is driving me crazy. I have googled for months and tried everything to figure out how this is done.

If I do the math and an 2 hour movie is 5GB and perfect, clear, sharp, then I should be able to have my 15 minute 4K move be about 650MB. The only way I found to render my movie out and get it to 650MB is to set the bitrate to 3000 KB/s and then it looks like crap. Can someone please tell me how this is done.

Thank you, John

open your goal clip in vlc and have a look at the codec information. (right click anywhere in the video, tools, codec information, should give you anything you need)

my guess would be h.265 as the magic behind it.

Sent from my Pixel 3 XL using Tapatalk

Ya, I have done all that. Some of the HD movies are as low as 850 KB/s and they look perfect, how are they doing this?

And I always do my videos/movies with h.265, that helps a little but still not even close.

My guess would be that the original source material is much cleaner (and larger) than your Sony generates and that they can use custom compressors that have capabilities beyond what FFMPEG, BMD, Apple, or anyone else can provide. Maybe they use off-the-shelf compressors and just really, really know what settings to tweak, but they can also afford to have a 300-machine render farm do a 4-pass compression with custom pre-processing and post-processing that's totally proprietary.

To the extent that it's simple pre-processing (blur by this much, sharpen by that much, etc, etc) I hope some insider is willing to share. So bump this thread.

I think the key is multiple pass encoding. The encoder analyses the video in preliminary passes, logging which macro-blocks require more bitrate and which ones can do with less. The final pass uses this info to compress the video in the most efficient way.

So on the multipass mentioned, it this something I would/can do with Devinci or are you suggestion a third party app, and if so what? I have myFFmpeg and Handbrake.

Basically what would be the best workflow to try?

"Real movies" are already processed/compressed in ways you won't be able to equal on your desktop, and you get the benefit of that first round when you add additional compression.

For your own footage, along with multi-pass, you'll probably want to add sharpening. And don't use Resolve for h.264/5 exports. Take the program into something like Handbrake, which offers more control and better results in smaller files. The "best workflow" there is what you arrive at after running tests.

Multipass is only useful if you need to hit a specific target size. It doesn’t guarantee increased quality compared to a single pass with similar bit rate.

Clean source material is the most important thing. Then use a really good encoder and learn its settings. x264 and x265 are good encoders for H.264/H.265. Both are available in FFmpeg and Handbrake.

Thanks, and right, I don't need to increase quality, I just need to try to preserve it best as possible and make the files as small as possible. I have some family and friends that have very slow internet speed, less then 1MB down and I just want them to be able to view the movie as clear as possible.

I would just upload the master file to YouTube. Then the client can decide the appropriate quality for the connection. And the user can override it if needed.

roger.magnusson wrote:I would just upload the master file to YouTube. Then the client can decide the appropriate quality for the connection. And the user can override it if needed.

the problem with that is then I know they will get 360 for there res and I don't want that.

(I know the data that I am going to give you in Spanish language. I hope your translation will be successful).

As I think Mr. Gary Hango says, it is essential that the encoding is VBR, which increases the bitrate in complex scenes and decreases it in simple ones.

In addition, I would always have to choose the HEVC (H.265) format because it gives more quality and occupies half the size.

Regarding the bitrate (counting only the video) i think that 4000 kb/s is enough to create a 4K HDR 10 bit video.

In the Davinci Resolve Studio version that I use (due to my simple GPU) I can only select Rate Control "VBR" by choosing the "NVIDIA" encoder, but I am limited to the "H.264" codec.

You will tell me if you can do it with the "H.265" codec, and if the resulting quality is correct.

Yes I know all this. but thanks. FYI I been a photographer for 43 and doing Video and sound a long, long time... when Devici renders my 4k at anything below 12000 kb/s it looks terrible. but with handbrake and 2 passes I can go down to 3000 kb/s and it looks GREAT.

I got it. figured out. the original video I out put was 5gb, I re rendered in handbrake and its not 300MB and it looks way better then the 2gb default YouTube video.

But now I just found a bug or something in 17,06 beta. its rendering all my audio in mono! its says its in stereo but the final file is mono.

So I found the bug, I will repost the info in the right place.
select H.265 in the master render settings it says audio will be in stereo but it outputs only mono.

Use FFMPEG exporter with it's x.265 encoder. There is a thread that explains how to add it to Resolve's output presets. Search for it.

Sent from my Mi 9T using Tapatalk

You already got good advice, like using FFmpeg instead of DR's encoding.
But to explain why you'll never get what Hollywood or commercial footage looks like, you need to understand that your Sony 'Pro' camera (you didn't mention the model) is probably using much more compression than true professional cameras. If it's H.264/265 it's using a codec that was never made for acquisition, but for distribution, i.e. the last step in the chain. If that's applied in the first place, for recording, it'll compromise the quality of the outcome whatever you try. Resolve has to decode your original footage into uncompressed 32-bit float RGB before you work with it. But in the end, you'll have two more GOP compressions if you output H.264/265 and YT is doing it once more. We once tested 10 generations of GOP compression vs. 10 generations of a mezzanine codec (the original clip was low-compression R3D). H.264 looked like an avant-garde manga, while the mezzanine codec was hardly discernable visually.

Cascading GOP compression is BAD. Period.
You may be better of with external recording for your Sony, if my assumption is right.

Mario Kalogjera wrote:Use FFMPEG exporter with it's x.265 encoder. There is a thread that explains how to add it to Resolve's output presets. Search for it.

Sent from my Mi 9T using Tapatalk

Are you talking about this link? viewtopic.php?f=21&t=73938#p641956 doesn't look like anyone has actually gotten it to work. yes that would be great to have because FFMPEG really works so mush better when you are trying to make the video small and keep it looking good.

Handbrake worked so nice on my video and what I was trying to do. I got my 15 minute video down to 280mb's, I converted to HD for the size but it looks better then it did with when I picked the DR's default YouTube 4K settings and it was 2GB's--amazing. But yes it would be so nice to have all these options in one package.

Until it works: Handbrake is based on the same foundation, there will be no difference in quality.
BTW, Shutter Encoder has a nice feature: "work during inactivity".
You can drop a list of clips to encode and it will take care of them when you take a break.

ajzapf wrote:Yes I know all this. but thanks. FYI I been a photographer for 43 and doing Video and sound a long, long time... when Devici renders my 4k at anything below 12000 kb/s it looks terrible. but with handbrake and 2 passes I can go down to 3000 kb/s and it looks GREAT.

I got it. figured out. the original video I out put was 5gb, I re rendered in handbrake and its not 300MB and it looks way better then the 2gb default YouTube video.

But now I just found a bug or something in 17,06 beta. its rendering all my audio in mono! its says its in stereo but the final file is mono.

In practice I have found the DVR exports in H.264 to be subjectively very "unsatisfactory" and I think this has been reported by other people as well judging by Internet search results. I always export DnxHR-444 from DVR for a master copy then use HandBrake to convert to H.264/265 using two-pass. The advantage of this method is that with all the overhead of color grading + effects + NR the DVR export render is very slow at 2-5fps and taking upwards of an hour typically depending on the length of the timeline - with 25% cpu utilization due to my GPU bottleneck, and it makes sense to have one archive quality master format given the excruciatingly slow render process. Handbrake on the other hand utilizes all cores and does the conversion job from DnxHR-444 into the target format in a few minutes for same video whenever I want it.

PS: I am even able to obtain very good results with Handbrake Noise Reduction+Sharpen during transcode if I skip NR in DVR, but I prefer to observe a few frames of every clip with NR+Sharpen turned on within DVR in order to tweak things for a WYSIWYG quality assurance before the painfully slow export.

ajzapf wrote:

Mario Kalogjera wrote:Use FFMPEG exporter with it's x.265 encoder. There is a thread that explains how to add it to Resolve's output presets. Search for it.

Sent from my Mi 9T using Tapatalk

Are you talking about this link? viewtopic.php?f=21&t=73938#p641956 doesn't look like anyone has actually gotten it to work. yes that would be great to have because FFMPEG really works so mush better when you are trying to make the video small and keep it looking good.

Handbrake worked so nice on my video and what I was trying to do. I got my 15 minute video down to 280mb's, I converted to HD for the size but it looks better then it did with when I picked the DR's default YouTube 4K settings and it was 2GB's--amazing. But yes it would be so nice to have all these options in one package.

No, this link: https://r.tapatalk.com/shareLink/topic? ... source=app

However, my mistake, it's R17 and x.264 only.
Hybrid by Selur FFMPEG GUI is full of options, just not for the faint-hearted.

Sent from my Mi 9T using Tapatalk

Mario Kalogjera wrote:

ajzapf wrote:

Mario Kalogjera wrote:Use FFMPEG exporter with it's x.265 encoder. There is a thread that explains how to add it to Resolve's output presets. Search for it.

Sent from my Mi 9T using Tapatalk

Are you talking about this link? viewtopic.php?f=21&t=73938#p641956 doesn't look like anyone has actually gotten it to work. yes that would be great to have because FFMPEG really works so mush better when you are trying to make the video small and keep it looking good.

Handbrake worked so nice on my video and what I was trying to do. I got my 15 minute video down to 280mb's, I converted to HD for the size but it looks better then it did with when I picked the DR's default YouTube 4K settings and it was 2GB's--amazing. But yes it would be so nice to have all these options in one package.

No, this link: https://r.tapatalk.com/shareLink/topic? ... source=app

However, my mistake, it's R17 and x.264 only.
Hybrid by Selur FFMPEG GUI is full of options, just not for the faint-hearted.

Sent from my Mi 9T using Tapatalk

It would be so nice to have some sort of FFMPEG H.265 in DR -- "With the estimate file size" like in Premiere--I really Really want that. Plus support for AISO!

I am still in the middle I love DR and keep trying to use it but I also like a lot of things in Adobe Premiere Pro! I hope someday DR will add the things I have mentioned.