Not sure what that image is showing other than a cropped image properly scaled on preview.

My question was really about under-the-hood processing, affirming the optimal things are happening on the render.

I've occassionally seen an image scaled to output resolution first and then cropped to fit, essentially scaling at that point instead of just cropping into the 4k source image.

Pan/ crop works at source media resolution not project so should be fine for this application.

 

The assumption:

• Rendering that project to a 1080p AVC .mp4 file will have zero quality loss from the original source media, since Vegas will be smart enough to know that it's just cropping into the image and not doing any scaling.

Is this accurate?

Pan/crop is the tool you use as people have said. I did a comparison between Vegas and Resolve and there's 2 problems with Vegas. I am seeing aliasing while zooming in, and the picture is softer than Resolve. I did not expect this so the zooms aren't perfectly identical but very similar.

2nd problem is Vegas is rendering at 2fps, there's nothing going on except zooming on on a high resolution picture, no fx or grading. This is the default 8bit processing mode encoding to Magix AVC hardware encoder.

Possible solutions for soft picture/aliasing, don't use MagixAVC hardware encoder, use 32bit float processing. I'll check that at another time

For final renders I usually use Voukoder x264 and output looks sharp to me (haven't done extensive tests as it looks good on my screens). It's slower though.

I usually do 4K to HD output and it's bordering on too sharp even with no sharpening added in camera or post.

For final renders I usually use Voukoder x264 and output looks sharp to me (haven't done extensive tests as it looks good on my screens). It's slower though.

I usually do 4K to HD output and it's bordering on too sharp even with no sharpening added in camera or post.

@RogerS Try with Voukoder the HEVC (NVIDIA NVENC) rendertemplate, which renders extremely fast and with good quality vs filesize. At least on my system, which is comparable with yours.

@3POINT and @RogerS  Are you using Voukoder because that's just what you're used to now, and at one point tried it BC of some problem you were having with native Vegas encoding/renders, or are there benefits today that you think make it superior to the built-in Vegas renders?

Asking because I've been happy with the results from Magix AVC renders (Internet HD 1080p 29.97 fps (AMD VCE) and also the older Sony AVC/MVC ones (Internet 1920x1080-30p). My target for those is Vimeo playback, but in my workflow I:
1. Render to the 1080p .mp4 file
2. IF Blu-ray is required, drop that .mp4 render into a timeline and render the Blu-ray from there (yes, 2nd gen render, but significantly faster than full re-render from source which often has noise reduction and other processing)
3. IF DVDs are required, similar to #2 but render to MainConcept MPEG-2 format sized to best fit the disc with highest bitrate I can

@RedRob-CandlelightProdctns

Posit:

"The assumption:

Rendering that project to a 1080p AVC .mp4 file will have zero quality loss from the original source media," 

I constructed three projects with identical Project Properties with different crop settings and rendered to the default Magix AVC/AAC render template with the following measured results.

Date: 2023/03/25  06:43:47  
Description: Crop Test Legacy 8 bit 2614.4 x 1470.6 Centered
Frames Processed: 330
Mean Squared Error: 3.296
Peak Signal to Noise Ratio: 43.008
Structural Similarity Metric: 0.8254

Date: 2023/03/25  07:30:15  
Description: Crop Test Legacy 8-bit 2688x1512 Centered
Frames Processed: 330
Mean Squared Error: 3.338
Peak Signal to Noise Ratio: 42.951
Structural Similarity Metric: 0.8234

Date: 2023/03/25  07:39:08  
Description: Crop Test Legacy 8 bit UHD to FHD
Frames Processed: 330
Mean Squared Error: 4.987
Peak Signal to Noise Ratio: 41.224
Structural Similarity Metric: 0.7655

Date: 2023/03/25  09:06:47  
Description: Crop Test Legacy 8 bit UHD to UHD Magix AVC/AAC
Frames Processed: 330
Mean Squared Error: 5.104
Peak Signal to Noise Ratio: 41.137
Structural Similarity Metric: 0.7519

Date: 2023/03/25  09:29:01  
Description: Crop Test Legacy 8 bit UHD to UHD Voukoder CRF22 GOP30 GOPMin15
Frames Processed: 330
Mean Squared Error: 5.318
Peak Signal to Noise Ratio: 40.92
Structural Similarity Metric: 0.7512

Typical Settings:

I was always curious if non-integer crop factors affect the end result. I discovered for this small sample that it does, but not negatively. The non-integer crop factor (2614.4 x 1470.6) was even better than the straight 3840x2160 to 1920x1080 render. Better, but not "zero quality loss".

Edit:

In an effort to determine the reduction in video quality from scaling, I thought it prudent to run tests of files with no scaling at all, that is 3840x2160 to 3840x2160. I also ran measurements of the default Magix AVC/AAC render template versus the Voukoder render template that I typically use. I appended the results to the end of the previous results measurements.

Conclusions:

• Most of the loss of quality comes from rendering to a lossy codec with a much smaller loss of quality contributed by crop factors.
• At the same bit rate, GOP and without B frames, Magix and Voukoder produce surprisingly similar results.

Playlist while this analysis was running:

... and for a slightly different groove...

Side Note: I sometimes feel like John the Baptist, a small voice crying out in the wilderness.

That event is pushed in (via Pan/Crop) 30% (image is larger)

Why? Assuming by "a 4K media event" you mean 3840x2160, the only zero-loss crop factor is exactly 2:1.

That event is pushed in (via Pan/Crop) 30% (image is larger)

Why? Assuming by "a 4K media event" you mean 3840x2160, the only zero-loss crop factor is exactly 2:1.

Are you asking why only pushed in 30%, when I presumably would have zero-loss if I cropped in 50%? Just a number I chose for the example, which if things work as I'd expect would not lose any pixels. I should be able to crop into that image up to 50% and at any amount, not lose anything since it's just cropping pixels and not scaling and dithering... at least, that's what I'd hope.

@3POINT and @RogerS  Are you using Voukoder because that's just what you're used to now, and at one point tried it BC of some problem you were having with native Vegas encoding/renders, or are there benefits today that you think make it superior to the built-in Vegas renders?

Asking because I've been happy with the results from Magix AVC renders (Internet HD 1080p 29.97 fps (AMD VCE) and also the older Sony AVC/MVC ones (Internet 1920x1080-30p). My target for those is Vimeo playback, but in my workflow I:
 

My experience (I always render my 2160p50 to 1080p50) that it makes a huge visible quality difference rendering with Voukoder instead of Magix AVC/HEVC. With Magix AVC/HEVC I see a lot of artefacts in regions with almost same colour and fine details (like water, grass, shadows etc).

So I render with Voukoder for it's superior quality against Magix. Also when using NVIDIA NVENC support with Voukoder, it's rendering much faster than using NVIDIA NVENC support with Magix.

It's Voukoder that I'm still editing with Vegas, Vegas without Voukoder, I probably would change my NLE...

Anything other than an exact common factor of the original involves some form of scaling, which is always lossy.

Vegas pan/crop uses bicubic scaling, which in the case of downscaling is about as good as any. But the relative pixel loss / interpolation is easy to calculate using basic ratio math, which is taught in every middle school, but rarely competently in the US.

No argument that Voukoder often gives better results, but they are not from scaling, which occurs upstream from the encoder.

I realize my thread was about maintaining quality while pushing into a 4K image for 1080p render -- but I have a related question (hopefully not steering off-base too much).

When I am previewing the 4K imagine in Vegas, it looks VERY soft to me in the Video Preview window regardless of what my quality setting is (preview/good/best, auto or full). Adding a sharpen filter of any kind seems to instantly make it look somewhat better, but still not the sharp image I'm certain the source is.

Anyone else ever experience that? I can't say I've done extensive testing with decode and project settings, but I have done some without difference.

@RedRob-CandlelightProdctns

 

Conclusions:

• Most of the loss of quality comes from rendering to a lossy codec with a much smaller loss of quality contributed by crop factors.
• At the same bit rate, GOP and without B frames, Magix and Voukoder produce surprisingly similar results.

@john_dennis Interesting indeedly! "to a lossy codec" -- would you say the Magix AVC codecs fall into that category, or was that a more general observation? Just wondering if I should invest any time to try a different renderer like Voukoder.

Also -- gotta admit -- I don't know how to interpret your quality/loss results (or how they are measured)... they look scientific and all, I just don't know what they mean!

 

Playlist while this analysis was running:

Sledgehammer.. yeeeeeaaaaasss!

@RedRob-CandlelightProdctns What is the size of your preview window when judging 4k?

@RedRob-CandlelightProdctns What is the size of your preview window when judging 4k?

@3POINT Project: 1920x1080x32 29.970 Preview: 960x540x32 29.970 Display: 158x893x32

Oh -- huh.. did I just answer my own question there?

Is Vegas scaling the 4K image down to Preview size, then up-scaling *that* up to the Display size?

That's the kind of thing I was concerned might happen with the cropping and rendering -- an inefficient sequence that actually worsens the image.

To judge a 1920x1080 project in the preview is only possible on an external monitor with a 1920x1080 display in best/full. Everything less is just guessing.

@3POINT Makes sense. I'm pretty sure I've pressed the "preview on external monitor" button at least once or twice when checking this on one of my 3 monitors... but I'll pay more attention next time.

 

Possible solutions for soft picture/aliasing, don't use MagixAVC hardware encoder, use 32bit float processing. I'll check that at another time

I have updated information relating to this

Vegas pan/crop uses bicubic scaling, which in the case of downscaling is about as good as any. But the relative pixel loss / interpolation is easy to calculate using basic ratio math, which is taught in every middle school, but rarely competently in the US.

No argument that Voukoder often gives better results, but they are not from scaling, which occurs upstream from the encoder.

@Musicvid Actually I found the fault to be the Vegas scaling, using Voukoder's internal default scaling works much better, as good as resolve, no aliasing, and looks less soft then Vegas's internal scaling.

It's why it's important to do these comparisons. Based on my result if you want the sharpest scaling with least artifacts you have to use Voukoder for scaling. And the worst part of this, nobody even knew that.

Curious -- Which part of the image are you looking at?

Also curious if the results would be the same if using a video instead of still image. The *obvious* answer *should* be "yes" -- since the only thing changing here is the encoding. But since I've had issues with Vegas not releasing memory when processing still images and video acceleration enabled, it makes me wonder.. and as you pointed out, the scaling appears to actually vary based on the encoder.. so.. maybe under the hood something isn't as obvious as we'd think?

The aliasing as i'm calling it, maybe not the right term, is the strobing of the black lines around the coast, There's tiny amounts of that with Resolve and Vegas via Voukoder scaling. The softness I"m mainly looking at the names when zoomed in. To me the Vegas (internal scaling) looks softer. Note that this isn't a great image because it's a little soft it'self and has artifacts due to upscaling it 2x (outside of Vegas)

I think this would be a problem for video too, but I"M zooming in 10X to this photo, for your work, maybe doesn't show. I'll look more into this in the future

Just installed Voukoder. Did one small 33 second render test of my project, using the default Voukoder template/setting and Magix AVC w/AMD render and the 2nd.

My subjective observation --- they looked and sounded nearly identical to my senses.

Interestingly though, the Voukoder defaults were much lower bitrate for that excellent result.

• Voukoder -- render time: 1:27.22 size: 18,729 KB Bitrate: 4,447 kb/s, audio 128 kb/s
• Magix AVCw/AMD -- render time: 1:15.17 (Radeon GPU @ about 20% during render) size: 49,928 KB Bitrate: 12Mb, audio: 192 kb/s (my preference)

Honestly, a 1080p bitrate under 8,000 I'd think would be more lossy than I'd like normally, but this still did look quite nice. Need to play with how to change Voukoder settings and look into ideal ones as well, noting quality is more important to me than render times (to some extent)

@RedRob-CandlelightProdctns It's a different workflow if you want voukoder to do the scaling. You feed it your source resolution, so 4K for you, then in voukoder it scales the image. It's in the filter settings

This is potentially a slower process, due to voukoder having to ingest 4K and convert it to 1080P. Even if you don't find any differences for your work, voukoder will give consistent results via x264 or hardware encoding, whereas MagixAVC Hardware has major problems with a lot of video, and MagixAVC has more limited problems with some video.