@Banani

Hi and welcome to the forum

. . . . After the export the video gets very blurry, sharpness is gone and quality just reduced alot . . . Video: H.264; 2704x1520p; 29.97 frames/s; VBR 6000 kbit/s; hardware encoding . . . .

That is not surprising - the bitrate setting for the export is too low for 2.7K video.

The VBR Average and Maximum bitrates for h.264 video 2.7K (1520p) export needs to be 32000 and 60000 kbit/s respectively.

HTH

John EB
Forum Moderator

Thank you very much for the fast reply John.

I have tried some other bitrates before you replied and it is a pretty big differince. It still is not as sharp as the original raw file, but I think that would be impossible. The 32000 and 60000 kbps would be average and maximum then? I putted a screenshot of my advanced options before I export the file. You think this would be the best I can get out of it?

By the way, in the original file, the kbps is 45000.

Thanks in advance

@Banani

Hi.

One thing I noticed is the GOP length. It has to be as close the the value of the frame rate as possible, so if you are recording at 29.94 fps you should set the GOP length to 30.

Ray.

@Banani

Hi.

One thing I noticed is the GOP length. It has to be as close the the value of the frame rate as possible, so if you are recording at 29.94 fps you should set the GOP length to 30.

Ray.

 

Ok thanks, I will also put that option on.

Belgium is a 50Hz PAL area, for 2,7K i can use 240, 120, 60, 30 and 24 fps. Should I consider using another fps ratio next time?

@Banani@CubeAce

Hi

. . . . The 32000 and 60000 kbps would be average and maximum then? . . . .

Correct.

. . . . the original file, the kbps is 45000. . . .

Use that value for the Average bitrate, with a Maximum of 60000.

. . . . It has to be as close the the value of the frame rate as possible . . .

Actually no - the Youtube specification specifies a GOP length half the framerate - in this case export at 30fps to match the source video with a GOP length of 15. This give maximum quality vs file size for the YT converters to work with

Youtube are going to alter the bitrate and GOP lengths on upload as it creates various versions of video resolution down to as low as 240p depending on a viewers Internet connection speed.

John EB

@Banani@CubeAce

Hi

. . . . The 32000 and 60000 kbps would be average and maximum then? . . . .

Correct.

. . . . the original file, the kbps is 45000. . . .

Use that value for the Average bitrate, with a Maximum of 60000.

. . . . It has to be as close the the value of the frame rate as possible . . .

Actually no - the Youtube specification specifies a GOP length half the framerate - in this case export at 30fps to match the source video with a GOP length of 15. This give maximum quality vs file size for the YT converters to work with

Youtube are going to alter the bitrate and GOP lengths on upload as it creates various versions of video resolution down to as low as 240p depending on a viewers Internet connection speed.

John EB

 

Ok, so with other resolutions, I would always need to put the original size in the average size option. But how do I know what maximum to put in then? Most of my video's still are in 1920 x 1080p (60fps). But in the original file the kbps is also 45000. And before Magix would always put in 6000 and 8000 for the 1080p videos.

For the GOP length, Should I always put in half of it as the original FPS? Like 30 GOP if I use 60FPS?

Thanks in advance :)

@Banani

Hi

. . . . Most of my video's still are in 1920 x 1080p (60fps). But in the original file the kbps is also 45000. And before Magix would always put in 6000 and 8000 for the 1080p videos . . . .

Which preset are you selecting - from the values you are quoting it sounds like you are selecting HEVC export not MP4?

As you can see below - the default 1080p 59.97 fps bitrates are much higher than those you are quoting.

Check the Display all option circled in red in the image above to see all the mp4 export options

In the above image there are some presets you will not have - these are custom ones I have created.

John EB

@Banani@Banani

Hi.

After John's explanation I looked at the specs again in the manual as I thought I had it correct originally but I had misunderstood the GOP structure. So I've been doing that bit wrong myself although I've always been happy with the exported files. So I'm trying the manual's recommendations now, of using a maximum GOP length of 15 for PAL conversions or 18 for NTSC if I'm reading it correctly now, that is the maximum GOP lengths those formats can handle correctly.

I have now tried a GOP length of 15 for a 50fps project and can't say I notice any difference between that and the one I produced using a GOP length of 50. So I guess the answer to that one would be 15 or 18 depending on the format of PAL or NTSC.

MEPs own base bitrate settings for given resolution / frame rates always seems fairly spot on for me quality wise. Going higher often just gives rise to larger finished file sizes with no appreciable increase in perceived quality of the final export. You will notice a degradation of quality if you go below the set limits in my experience though. I also personally find it best to use the 'Best' option in the drop down box as you have done. The manual also suggests for best quality to allow two passes rather than one although it will double the export time.

Ray.

@CubeAce

Hi Ray

GOP length is not a fixed in stone setting - different situations require different values.

For example:

• Direct playback of video on a computer or TV with a GOP value of half the frame rate is helpful in attaining good quality, all other parameters being equal and larger file sizes are not a restriction.
   
• Playback from disc, DVD or BD - the relevant standards rule.
   
• Video streaming sites eg You-tube, Vimeo etc modify videos you upload to provide the best combination of video size, bitrate and streamability with respect to the viewers Internet connection speed and their requirement specs reflect the starting quality they need.
   
• The bigger headache is streaming from a website (commercial or personal) ie not YT, Vimeo or other similar sites.
  
  Smoothness of the stream for the viewer is essential and an increased GOP length can help, along with offering several different resolutions and, in some cases different formats eg mp4, webm, and wmv (Windows devices only) can be a bit of a nightmare to administer, fortunately mp4 is now the most universal format to host.
   

After a lot of testing and analysing other streaming services content, I now use a GOP length of 100 for streaming 1080p 25fps mp4 video from a hosted website.

. . . . MEPs own base bitrate settings for given resolution / frame rates always seems fairly spot on for me quality wise . . . .

They are optimised to provide a good balance between file size and quality - you may have seen me recommending many times to never change the defaults for export.

HTH

John EB

I have found that, for whatever reason with MEP, it helps to set the sharpening up a little in the generic movie settings. Someone else here (John EB?) had mentioned that they do this too.

@Banani, @pmikep

Hi

. . . . Someone else here (John EB?) had mentioned that they do this too. . . . .

pmikep is correct, I do advise adding some overall sharpening using the Movie Effects, Sharpness tab - I use a setting of 30 for 4K and 1080p.

If you go to high with the sharpness setting you will begin to notice odd colours on edges or a 'speckling' effect.

John EB

Any speculation as to why this is? Is MEP purposely softening video as a Marketing thing, to make them look "better"? One would think that the video that goes in would be the video that comes out.

@pmikep

Hi

. . . . Is MEP purposely softening video as a Marketing thing, to make them look "better"? . . . .

If they wanted to make themselves look better sharpening would be automatically applied on export without you knowing - think of the advertising line for that one!

On a serious note, digital images are inherently softer then you would think, after compression in the camera, decompression in a video editor - any make, recompression on export, then decompression on the playback device. At each compression stage you are throwing away data, less data means lowering quality, for the most part this is not noticeable as perceived quality is what matters and is 'an illusion' not matching real quality which requires software to analyse.

If you look at a commercial DVD frame by frame you will see a high degree of sharpening is applied to the image, as the resolution goes up the degree of sharpening required decreases.

The most important thing is that sharpening effect should be the last effect applied Apply it earlier and it may be undone by other effects applied after. Rarely should it be applied to a single video or image object.

John EB

@Banani

I have tried some other bitrates before you replied and it is a pretty big differince. It still is not as sharp as the original raw file, but I think that would be impossible. The 32000 and 60000 kbps would be average and maximum then? I putted a screenshot of my advanced options before I export the file. You think this would be the best I can get out of it?

One thing I notice is that Profile / Level setting is incorrect for the resolution used and the increase in bitrate you have applied. Try a higher level setting 4.2 or higher.

@CubeAce @johnebaker

GOP length is a trade off between quality and efficient compression, within a set bandwidth, and ease of editing of the output file. A short GOP means there are more I frames which contain all of the information / pixels of the original scene (albeit compressed with something similar to JPEG compression). The down side of this is that these frames use far more bandwidth (bits) than the predictive B and P frames in between, so too many I frames reduce the quality within the set bandwidth rather than increase it, if the target export bitrate is set too low.

Too long GOP's like the 100 frames used by John are fine for narrow bandwidth streaming applications ,giving maximum compression efficiency and quality if that is the final destination for the file, but are problematic for further editing where cuts occur between distant GOP boundaries ( I frames).

Magix export templates formerly used a GOP of half the FPS rate but now appear to use a GOP equivalent to the export's FPS. Half could improve quality if there is plenty of bitrate headroom.

Peter

This thread has been very informative, thanks for the discussion, guys. 👏

I was going to guess compression was the culprit. And then I thought I might try to export to AVI or something with little to no compression to see if there's any softening there. And then I thought "I'll wait for Ray to try it." 🙂

OTOH, I don't notice softening of video when I make videos of the same type using competing software. Hmm... are they doing auto sharpening?

@pmikep @johnebaker @Scenestealer @Banani @AAProds

Just when I thought I was getting a handle on everything! 😂

Back to the manual for me then where everything is as clear as mud to me particularly when Magix offers this statement at the end regarding settings for codecs.

I also found where I got the 'Make the Group Of Pictures setting the same as the frame rate' idea from.

and as well as finding that advice on other pages.

Which when looking at the earlier advice confused me at first as how can you have a GOP length of an odd number compared to the frame rate until I realised this would be the interval between I frames. Which if I'm reading correctly may introduce drift when GOP lengths become excessive.

That in turn left me more confused because if MediaInfo is not giving the GOP information in the original file (or copies for that matter) then is it readable to MEP and is that important? Assuming I frames are as 'pure' as one can get with a compressed image and is a reference point and has a set interval length within the GOP structure, then when read for conversion could the reading be taken from a B or P frame instead? I've no idea but it's a disturbing thought to me.

My main concern though is the Editing monitor within MEP. Unlike editing a stills image from a camera raw file I am looking at what?

If I am copying a file into MEP then how accurate a copy? Even if looking at the source file from another location and not using it in full screen on a second monitor it is being re-scaled to fit (even on a second monitor it does not show full screen with no boarders.) So how accurate an image am I getting to evaluate something like sharpness?

I seldom sharpen much. I don't think I have ever sharpened a whole video but maybe just some slow motion footage taken in camera as I know my cameras use a crop to the central portion of the camera sensor to keep data rates to a point that the recording media can handle. That can lead to a softer image compared to when the whole sensor is in use on my cameras. To data I have been more concerned with the compression rate of the export that seems to effect the export more as far as I can evaluate.

As John EB has explained, sharpening can produce a noise. I find it akin to shooting at a higher ISO where it is more noticeable in shadow areas. It is the main reason I tend not to use sharpening if I feel I don't need it.

So up to a point I'm only using the editing screen to look at effects lengths and try to work out what I think the final file will look like. I don't make too many judgements on things like sharpness or contrast.

What I do is play the finished file against the original file or files and select frames at different points to see what differences there are. Looking where both the finest details for alterations, the shadows for noise, and at the areas with large bland areas for blocking or banding.

At the end of the day we all might be looking for slightly different things in our exports due to personal needs that may also be influenced by what we are shooting footage with that may also play a part in our decision making, and that I think may be the only sensible way of looking at it IMHO.

That also excludes what we are looking at in the way of monitors. I'm sure some here have much better monitors than myself and it may be within the realms of possibility that some can see defects I (or others) can't.

Ray.

@CubeAce

Hi Ray

Welcome to the 'rabbit hole' of video export settings 😂.

On a more serious the important parameters are:

Resolution, Average and Maximum Bitrates and GOP length and structure

Matrix quantisation is a heavy mathematical concept to get the hang of - I have a vague understanding of how it works and I am leaving it at that.

MediaInfo does give you 2 GOP parameters that are worth noting ie:

• Format settings, GOP     : M=3, N=100

M = the distance between two anchor frames (I or P)

N = the distance between two full images (I-frames)

MEP/VPX also give you the GOP structure as shown below - this is my web video preset

Putting the two together:

IBBP + M=3, N=100

The encoder does the following:

1. Start GOP with an I frame ( full image)
    
2. insert B and P frames in the sequence specified by the GOP structure, inserting a P frame every 3 frames (M)
    
3. on reaching the 100th frame (N) go back to step 1.

On a side note, there were some video cameras which started a GOP with a P frame (black) followed by the first I frame, if you look at the settings in MEP/VPX you will see there is an option to automatically remove these.

HTH

John EB

@pmikep @johnebaker @Scenestealer @Banani @AAProds

Hi John.

Thank you for the additional info. A question of me not seeing the wood for the trees. Again.

So there are reference points MEP can use to pick up on. That's one less worry then.

So I've looked at the media info from both the cameras I use, as to date I haven't really mixed that much footage from both into one project.

Apart from I will have problems with only being able to shoot at lower frame rates if I shoot 4K with my DSLR, the other is it is using mov. as a container.

both from my action cam at different frame rates

And one from my DSLR

My recordings from my action cam I had set to use MP4 although that too could be set to mov and use the same frame rates if needed. Would it make a real or noticeable difference beyond that of sensor size capabilities?

Both are using their full 20 megapixels (allowing for crop differences between ratios used). Both crop to the centre of the sensor for slow motion. The DSLR sensor output always looks softer down to sensor size / field of view / photo sensor size and aperture differences and obviously the DSLR has better noise control for a given ISO. But now I notice a difference in GOP structure as well between the cameras. Oddly to my mind the GOP structure does not change with frame rate on either camera.

So my next question will be what happens if I mix the footage produced from both cameras from the GOP perspective? I know there will be other issues as well that could end up with the footage not being seamless but just trying to get the best conversion from a final export.

@CubeAce

Hi Ray

. . . . The DSLR sensor output always looks softer down to sensor size / field of view / photo sensor size . . . .

If the lens is a stock zoom then it may be causing the softness.

I have a Canon DSLR and the stock zoom lens is 'soft', this became very apparent when I purchased an EF 70-200mm f/4L lens.

Using a DSLR as a video camera is my pet peev.

. . . . my next question will be what happens if I mix the footage produced from both cameras from the GOP perspective? . . . .

WRT to the GOP structure - no need to be concerned the GOP lengths are short enough not to fall foul of the long GOP editing issue @Scenestealer mentioned above.

However, from a different perspective, with different frame rates I opt to use the same frame rate for both my Sony cameras this avoids the potential issue from Pull Down when changing frame rates.

The DJI data - Format settings : 1 Frame and M=1 N=30 - looks like it is using I frames only and compressing them.

. . . . Oddly to my mind the GOP structure does not change with frame rate on either camera . . .

The internal structure sequence is independent of frame rate, hence you can have 50 fps with a GOP N = 30.

As far as the container goes MP4 or MOV, if the internal codecs are identical it does not matter which to use.

John EB

@johnebaker

Hi John.

I'm not sure what you mean by a stock lens but I use the the holy trinity set of lenses 14-24mm, 24-70 mm, 70-200m, (all f2.8 constant aperture) along with a few f1.4 prime lenses. None of which are soft although the 50mm f1.4 is slower to focus than the others. Also, as they are all full frame lenses working on an APS-C body, (for video) the extreme outer edges of the lenses are not in use. I don't often use those lenses at those apertures as motor noise can be a problem if I can't get a mic far enough away from the lens (Due to focus motor noise) and most of the time I'm using apertures of between f4 to f7.1 depending on the focal length I'm working at. Too shallow a depth of field doesn't work too well with event shooting but going too narrow an aperture can bring other problems I'm sure you are aware of.

The only reason I had used MP4 over mov. was that MEP seemed to work smoother on my old PC. Since my upgrade I've not really tried mov so I'll try switching.

If I recorded 4K with the DSLR I'm limited to a top fps of 29.94fps. That's usable for most of what I do. If I record at 1080p then I have the full basic frame rates to choose from. Either way if I were shooting with both cameras for the same project I would use the same frame rate for both.

Constant 'I' frames never occurred to me but would explain a lot and pose more questions. So now I'm wondering about the second figure which seems pretty meaningless apart from setting the length of the structure.

Thank you for the additional info.

Ray.

@CubeAce @johnebaker @AAProds

The DJI data - Format settings : 1 Frame and M=1 N=30 - looks like it is using I frames only and compressing them.

No, It appears to be just not using B frames, only I & P frames, hence a GOP structure that looks like IPPPP......PPPI. This may be why the DJI uses a generous 80Mbps or 100Mbps data rate for recording, due to the less efficient nature of the P fames compared to B frames. It may be also be for less processing when it creates the live feed for the transmission to the controller's monitor??

@CubeAce

I am interested as to how your DSLR uses the sensor when shooting video as opposed to still frames? You mentioned it cropping for high speed but does this change the angle of view for the same lens focal length setting and does this occur when reducing the resolution to 4K or HD?

Re your concerns about assessing sharpness in MEP with a smaller resolution preview window - I use the enlargement function, say 200 - 400%, from the hamburger menu top left of the monitor and place a copy of the clip on a track below the clip, apply the sharpening and then toggle the solo to compare with and without the effect at an enlarged pixel level. This is similar to the useful little crop that appears in Photoshop when using the sharpening tools.

Peter

@Scenestealer

Hi Peter

. . . . It appears to be just not using B frames, only I & P frames . . . .

Thanks for the correction - think I had a brain fade there, particularly as in the previous comment is said:

. . . M = the distance between two anchor frames (I or P) . . . .

How embarrassing 😱

John EB