Resolve 17.4 Video file size far to big

Hey,

after installing/updating to Davinci Resolve 17.4, I noticed the files size in the delivered mp4 (youtube) Format have gone up roughly 10 times to before.

Where a video of approx 8 to 10 minutes had a delivered filesize as a mp4 youtube format of 400MB to 650MB.
With 17.4 the file size is now 5GB

What happend here? My workflow, also in delivery, hasn`t changed.

Update to 17.4.1 didin`t change this.

What happend here, that the file size after rendering went up 10times the size as before?

I guess I am the only person with this issue, so I guess I will have to figure it out by myself.

I would say it probably has to do with the changes they announced in the update notes.

Codecs and File I/O
• Improved encode settings for YouTube render preset.
• Option to export a timeline marker color as YouTube video chapters.

I would say it probably has to do with the changes they announced in the update notes.

Codecs and File I/O
• Improved encode settings for YouTube render preset.
• Option to export a timeline marker color as YouTube video chapters.

10 times the previous file size is not really an enhancement, I would say.
Then the release notes should also state the increase of the file size.

To be honest, taking the changes made in the previous versions and the issues that have come with the newer versions, I am actually starting to regret buying the studio version.
DR is a really good software, but it has severe issues since version 17, that have only got worse. sorry to say that.

Compare the data rate of the new files vs. the old files you have created. It's likely that the new files have higher bandwidth which you can adjust in the Deliver page.

Just did a test (smallish file). I used either H264 or H265 from PROres worked just fine. Must be a setting you have somewhere. My 5900 and 6700 combo doing 4K is not showing any issues.

I havnˋt changed any obvious settings and That is what ticks me off on this.

Just did a test (smallish file). I used either H264 or H265 from PROres worked just fine. Must be a setting you have somewhere. My 5900 and 6700 combo doing 4K is not showing any issues.

What is the filesize after rendering?

The rendered 4K file is 6.64 MB in H264, Original files was 39.1 MB's. This is using 14.1 studio.

The rendered file is 6.64 MB in H264, Original files was 39.1 MB's.

I will gute H.264 a try

I consider the H.264 and 265 export encoders in resolve flawed and essentially unusable. I personally export to DNxHR HQ and use handbrake for H.265 conversion. This is still the case after I purchased Studio and got access to the Nvenc h.265 export in resolve.

Flaws with resolve exports.

1 Low-Best presets for h.265 are unusable. 'Best' gives like 20 mbit for 4k and shows horrible compression artifacts.

2 VBR encoding in resolve has zero documentation and cryptic descriptions of mandatory variables.


3. No high quality CPU based encoder for h.265, only Nevec which is proven to provide lower quality files at equivalent (low) bit rates to CPU based encoding (X265).

[quote="ZRGARDNE"]I consider the H.264 and 265 export encoders in resolve flawed and essentially unusable. I personally export to DNxHR HQ and use handbrake for H.265 conversion. This is still the case after I purchased Studio and got access to the Nvenc h.265 export in resolve.

Flaws with resolve exports.

1 Low-Best presets for h.265 are unusable. 'Best' gives like 20 mbit for 4k and shows horrible compression artifacts.

I use the supplied AMD convertor for H.265 and it's 40 mbit and ok for for online stuff.

You don't need to use the presets. Pick custom and make your own quality / size choices.

You don't need to use the presets. Pick custom and make your own quality / size choices.

I know. But I also am of the opinion, the presets are Set for best quality and performance for what it is set to. So in case of YT the Settings should be best for that use case.
In the file size however increases by almost 10 times the size before, then is a Bit out of range.

Itˋs Not only the files are bigger, this also means Upload time is longer, Storage Space (in corporate area this Van be crucial) bigger and more expensive, time on the Video plattforms is longer to geht the Video up and ready After upload.

But that's the thing. Somebody still on dial up using a small HDD is going to have a different idea of ideal filesize than somebody with a highspeed gigabit connection and virtually unlimited storage.

i had same issue et got back to 17.3.2

This isn't a solution, but a really decent workaround (for PC users). Voukoder is Codec plugin for Resolve -

Note: Make sure to download the "App" and the "Connector" for Resolve.

- Install the "app" and install the "connector" plugin.
- Launch Resolve.
- In the Deliver Page, select Voukoder from the pulldown.
- You can configure Voukoder by clicking "Configure Voukoder" under the plugin settings.
- Pick Output Codec, settings, enjoy.

https://www.voukoder.org/

a video of approx 8 to 10 minutes had a delivered filesize as a mp4 youtube format of 400MB to 650MB.
With 17.4 the file size is now 5GB

My (wild) guess is that somehow your export preset changed from VBR (variable bit rate) to CFR (constant frame and bit rate). CFR gives you better quality, but 10x file sizes.

a video of approx 8 to 10 minutes had a delivered filesize as a mp4 youtube format of 400MB to 650MB.
With 17.4 the file size is now 5GB

My (wild) guess is that somehow your export preset changed from VBR (variable bit rate) to CFR (constant frame and bit rate). CFR gives you better quality, but 10x file sizes.

That Sounds plausibel, will Check um on that.

Since there are no detailed settings available for YT encoding:

YT render settings.png (22.69 KiB) Viewed 16385 times

I assume that the settings are used from the H.264 / H.265 Master

YT render settings Master.png (37.64 KiB) Viewed 16385 times

as you can see the default ist Quality Best which leads to extremly large files !!
switch it to medium or better to restrict to, which provides great quality using a variable bitrate. For FHD on YT 16.000 may be a good value.

I will give that a try, as soon as I edit my next video and you all know.

Thank`s for the help so far

So, I tested it this morning dropping the quality to medium, variable bitrate, 16000 for YT and the file size is back to where it was before.

That means in the end, with the 17.4 update the default settings in the YT preset was increased to high quality and constant bitrate.

Increasing the quality is a good thing, but a detailed information, what was changed, would be nice aswell. There users like me out there, that are not professional video editors and don`t have the in deptch knowlege about such things.

anyway, thank´s for the help provided.

On my recent project (delivered yesterday) I just took the "H.264 Master" preset as a starting point.

Actual settings were:

1. Encoder: NVIDIA
2. Quality: Restrict to 80000 Kb/s (not "Best"!)
3. Encoding profile: High
4. Key Frames: Every 14 frames
5. Rate Control: VBR High Quality
6. Lookahead: 16 frames
7. Entropy Coding Mode: CABAC
8. ...everything else left unchanged...

"Format" (actually, which kind of container to use) may be MP4 or MOV, it doesn't matter.

For my FHD 1920*1080 29.97 fps timeline, this gave me some appx. 36 Mbps H.264 video bandwidth in the delivered file, actually almost the same as what my camera provides while shooting FHD 29.97 4:2:0.

The quality of the delivered result was quite satisfactory. The client was happy.

For my FHD 1920*1080 29.97 fps timeline, this gave me some appx. 36 Mbps H.264 video bandwidth

Not related to Resolve, but you never want to upload 1080p to YT. Using 4k it will get better treatment by its compression algorithm.

So even if you shot 1080p, upscale to 4k for YT upload.

Not related to Resolve, but you never want to upload 1080p to YT. Using 4k it will get better treatment by its compression algorithm.

So even if you shot 1080p, upscale to 4k for YT upload.

Thank you, my friend, I do know this hint already. But as I already mentioned, I do exactly what people ask me to; I was asked to provide FHD in H.264, I did it, no matter what they will do with the deliveries later.

Warmest regards, Andreas

Just to add more pieces to this puzzle.

As I mentioned before, with the manual limit set to 80 Mbps instead of "Best", I got the file of size 2.2 GB and ~36 Mbps bitrate.

Being curious, I re-rendered the very same timeline with the "Best" setting. I've got a file of 1.2 GB size, and ffprobe tells me it's 15861 kb/s video bitrate.

Again, I re-rendered the very same timeline with a manual setting of 18 Mbps (18000 Kbps). Now the file size is 1.3 GB and ffprobe tells me it's 17673 kb/s video bitrate.

So everything looks pretty consistent with one single question: why on the Earth does Blackmagick think that just 16 Mbps is the "Best" quality for H.264 encoding of FHD 29.97?! Isn't 36 Mbps (i.e.) the better "best"???

Well, for BluRay in HD you'd find data rates roughly between 14 and 20 mbps on commercial products.
So, it seems that BM assumes that 'Best' means 'good enough' here for delivery. You'd probably see no difference with higher rates. Of course, I'd go higher (or use Handbrake anyway) if it's for re-compression.

Well, both my cameras (Fuji X-T20 and X-E3 internal guts are identical) are doing FHD 1920x1080 at 36 Mbps H.264 no matter what frame rate is. So the final delivery at 36 Mbps makes sense - it looks like the deliveries retain exactly whatever the camera has acquired, and without a perceivable IQ loss.

Thus in my humble opinion, 36 Mbps is quite a fair bitrate for H.264 delivery which is intended to be a "master" copy. Pushing delivery bitrate higher than the camera acquisition bitrate looks like putting your $20 into a bigger vallet in the hope for some gain.

And yes, I definitely see the visual difference between 18 and 36 Mbps editions of my delivery, esp. on a large TV screen (on smartphone there isn't any noticeable difference, on TV the 18 Mbps editions looks soapy).

One side note. My Fujis are acquiring FHD at 36 Mbps H.264 no matter what frame rate is (also UHD is 100 Mbps again with no regard to fps), does this mean that shooting at fewer fps gives better video IQ, given the bitrate is constant?

For say UHD at 60 fps, compared to UHD at 24 fps, will IQ be better (less compression, better clarity) in the second case?

Warmest regards, Andreas

If the bitrate, which is also measured in units per second, is constant: yes, the image quality should be better. More data for every frame.

If the bitrate, which is also measured in units per second, is constant: yes, the image quality should be better. More data for every frame.

Thanks a lot for clarification. I've had this similar gut feeling (and visual impression, too, though the latter is my humble, subjective and biased personal opinion). Actually I don't know how to measure IQ in terms of numerical metrics, unfortunately.

One more interesting observation. I also own an old panny LX7, I love it for it's "IQ / bulkiness" ratio. Really tiny pocketable beast which is on par with Canon 20D by IQ. I wasn't using it often in the last two years, but now I decided to try it again as my third FHD camera occasionally, or as a main one when larger "professional" gear is forbidden on set.

The panny LX7 offers FHD in 2 options: AVHCD at 50 fps and 28 Mbps bitrate, and MP4 at 25 fps and 25 Mbps bitrate. So it looks like MP4 at 25 fps will provide somewhat better IQ compared to the "Blu Ray" like AVHCD, yes? Interesting hypothesis I will try it and compare these 2 options on a big TV (never thought of it before).

If it's really the older mp4 codec, it may be inferior nevertheless. But try by yourself.

Not related to Resolve, but you never want to upload 1080p to YT. Using 4k it will get better treatment by its compression algorithm.

So even if you shot 1080p, upscale to 4k for YT upload.

BTW this is probably a common misconception or urban legend.

This article (link!) clearly informs us, that only the UHD version of your upload is stored in VP9 by YouTube. So when you upload UHD, the UHD resolution edition will be stored and delivered in VP9, but deliveries at downscaled resolutions are to be transcoded by YouTube, then stored and delivered in H.264 again.

Right yesterday I tested it. I had a project with an FHD timeline, I changed the timeline to UHD, exported it into 80 Mbps UHD H.264 and uploaded it to YouTube. The only difference I have observed with regard to the FHD version was the vastly longer processing time. I can give you a link to watch, in case you are actually interested, to your pleasure, my friend.

So for an FHD project, I decided that upscaling to UHD probably makes no sense. Just upload your FHD in, say, 36 Mbps H.264, then YouTube will transcode it to some smaller bitrate H.264 under the hood anyway, so for FHD delivery, there should be no difference between upscaling and not, other than wasted time and disk space.

UPD. I've fixed the misprinted URL thus the edit of the comment appears

BTW this is probably a common misconception or urban legend.

No, people have done VMAF testing of the files you get back from YT with different settings

https://andsobegins.com/article/2020/09 ... uTube.html

Granted this guy used game footage, not regular camera footage. Not sure of that makes it better or worse in YT eyes.

1080p files of even 26mb got at quality score of 52, and most of his tests are around 60

4k files by comparison the lowest is 86 and some are 99

you see in their "“2160-1080 YT.vp9 VMAF" column 4k watched at 1080p has 88-97 quality, very good

So you will get a better experience, shooting 1080p, upscalling to 4k, export 60mb h.265, upload to YT, watch at 1080p. Than if you uploaded at 1080p


Another guy tests various bitrates 20-100mb for YT uploads with VMAF and found 60mb for 24-30fps footage is the point of diminishing returns. But I can't find this source for the life of me.

And it's all pretty strange from YT's side, since they unnecessarily generate incoming bandwidth, processing power, and waste of energy. Why can't they just encode 1080 at the same quality, if all we need is HD? Most viewers will hardly notice the difference on their cellphones.
Do they want to force us into 4K/UHD production? They make their money from ads, not from cameras or computers. Or did they silently invest into hardware manufacturers?

Sorry for the rant, but in my eyes this is about as sick as bitcoin mining.

And it's all pretty strange from YT's side, since they unnecessarily generate incoming bandwidth, processing power, and waste of energy.

Actually we just aren't aware of the logic which their product management is following to. Business "logic" is nothing close to a common sense.

First of all, I bet their player is checking the actual data rate and bandwidth for each connection to their servers. So if you are on 8 Mbps ADSL they will feed you a downgraded video bitrate (like 6 Mbps max). If the actual bandwidth between their SDN and your host is 100 Mbps, they will send you the stream at its full video bandwidth.

So probably they have some (massive!) statistics for how fast the connectivity of the majority of the viewers is. And they are adjusting their storage format accordingly. This is just my wild guess, of course.

The only meaningful outcome of my speculation is the following hypothesis:

1. YouTube will provide you with a video stream of a bitrate which is ingestible for your internet connection bandwidth.

2. This bitrate will never exceed some corporate-stated limit which is established internally and is not available to the public.

3. With the growth of the median of the individual connectivity bandwidth worldwide, this limit may rise but we all won't be informed.

Why would they store videos with a bandwidth exceeding 8 Mbps if 90% of viewers have their internet connection bitrate below 10 Mbps?

Do they want to force us into 4K/UHD production? They make their money from ads, not from cameras or computers.

You don't actually have to shoot or edit 4k. You can just upscale at export or even years later with ffmpeg. Its just to trick the YT algorithm

My GTX1650 4gb laptop could not do 4k. I would edit 4k gopro or drone footage on a 1080p timeline, export at 4k, clicking yes at the resolve warning 'are you sure you want to do this silly upscale?'

So you actually think that in comparison between

Scenario 1. FHD H.264 delivery and upload with excess bitrate and YouTube transcodes it into some lower (X Mbps) bitrate H.264, and

Scenario 2. FHD upscaling to UHD, uploading it, then YouTube transcodes it back into FHD w some lower bitrate H.264 (Y Mbps)

you get Y > X with observable difference?

If yes, then why?

So you actually think that in comparison between

Scenario 1. FHD H.264 delivery and upload with excess bitrate and YouTube transcodes it into some lower (X Mbps) bitrate H.264, and

Scenario 2. FHD upscaling to UHD, uploading it, then YouTube transcodes it back into FHD w some lower bitrate H.264 (Y Mbps)

you get Y > X with observable difference?

If yes, then why?

It's shown in the blog linked by ZRGARDNE that the youtube encoding pipeline ingests 2160P video differently to 1080P. Including superior downscaled 1080P from that upscaled and compressed version.

Between the two blog posts in the series, their first test clip has a VMAF of 57 for 1080P/YT_H.264, 65 for 1080P/YT_VP9, and 72 for 1080P->4K bicubic->1080P/YT_VP9. The 1440P version of the 4K upscale had a VMAF of 86 and the 4K upscale a VMAF of 89.

So you actually think that in comparison between

Scenario 1. FHD H.264 delivery and upload with excess bitrate and YouTube transcodes it into some lower (X Mbps) bitrate H.264, and

Scenario 2. FHD upscaling to UHD, uploading it, then YouTube transcodes it back into FHD w some lower bitrate H.264 (Y Mbps)

you get Y > X with observable difference?

If yes, then why?

It's shown in the blog linked by ZRGARDNE that the youtube encoding pipeline ingests 2160P video differently to 1080P. Including superior downscaled 1080P from that upscaled and compressed version.

Between the two blog posts in the series, their first test clip has a VMAF of 57 for 1080P/YT_H.264, 65 for 1080P/YT_VP9, and 72 for 1080P->4K bicubic->1080P/YT_VP9. The 1440P version of the 4K upscale had a VMAF of 86 and the 4K upscale a VMAF of 89.

To me, what's crazy is the fact that this "fact" is known for many years, Linus Tech Tips talked about it in one of their video...7 years ago, in 2015.

here :

. Especially @ 6:20

But I never saw anyone else really talk about it in the "mainstream" tech media, or in general since then.

When I started editing 3 years ago, I remembered this video from Linus Tech Tips, and I always exported my 1080p timeline to 4k.

Like other people said here, Youtube is using a better codec for anything higher than 1080p (even 1440p is fine for that, so if size and hardware is a problem for you, render your video at 2560x1440 instead of 4k and you will still benefit from the better codec used to encode the video).

I shoot in HD but always upscale to UHD for YT. I make railway videos which involve shots from moving trains with a lot of lineside foliage passing by. In HD you get very bad compression artifacts with the moving foliage, something that is almost totally eliminated with a UHD upload.
Camera originals are 50fps 35Mbps mp4, and rendered to UHD at 50fps 80Mbps mov.