Huffyuv Capture Sanity Check

[Xhumeka]

Hi folks - I've been using a cheap Elgato as a "first pass" capture on a large tape collection, and now I've started the task of recapturing important content using VirtualDub and huffyuv lossless codec, setup as per the guide here: https://www.digitalfaq.com/guides/video/capture-avi-virtualdub.htm

After having captured some clips using this method, I tried to render to .mp4 to make sure I was doing everything properly before spending countless hours continuing this process. I'm no expert and I'm trying to use Handbrake, not the recommended Hybrid (which I plan to learn eventually!) but I expected to at least do better than the basic Elgato capture!

But no matter which settings I play with in Handbrake, I can't for the life of me get footage to look better (or even as good) as the Elgato. I wondered if you folks could:

a) confirm my huffyuv capture looks OK

and

b) advise which settings/filters I should try using in handbrake to improve my render footage, or even better, provide a .mp4 rendered from my huffyuv file that shows it's possible to match elgato quality, then I can worry about figuring out how to do it myself later - I just want to make sure I'm not wasting my time doing something wrong in virtualdub/huffyuv.

Here is the link to an example of both the Elgato raw capture, as well as my huffyuv test capture... thanks to any who can offer assistance!

https://drive.google.com/drive/folders/1ylSWTdVsAXIMbYpkIkBwLwwKdwfPicy1?usp=sharing

(NOTE - both clips were split (not re-encoded) from larger files using ffmpeg... hopefully that doesn't matter, but if I need to re-capture smaller clips to provide as examples let me know)

[Bwaak]

Both use the same Elgato stick, but the "Elgato_Raw.mp4" is captured by Elgato software, while the "VirtualDub_Huffyuv.avi" is captured by VirtualDub? So, the Elgato stick can send unmutilated interlaced video, that is good.

The MP4 is deinterlaced into 30p, you've lost half of the temporal resolution and up to a half of spatial resolution. Ok for pre-2014 Youtube. It is not bad, actually, if you are willing to accept 30p, it seems that Elgato does some noise processing, and the sides are correctly cropped.

The AVI is OK. What do you want to do with it? Stash for better times? Edit in your favorite NLE? Save to a thumb drive and give to your parents? Upload to Youtube?

For watching on a computer or uploading to Youtube, do the following:

Video -> Filters -> Add... -> Deinterlace -> Interpolate using Yadif, Double frame rate, top field first
Video -> Filters -> Add... -> Crop -> crop to 704x480
Video -> Filters -> Add... -> Resize -> Aspect ratio: disabled, New size: absolute 960x720, Filter Mode: bicubic is ok.

You can add more filters to handle the noise, for example.

Audio -> Full Processing Mode
Audio -> Compression -> choose something to your liking, MP3 is a common choice, choose bitrate to your liking.
Video -> Compression -> x.264 8-bit -> Configure -> SAR width: 1, SAR height: 1, Quantizer: 20 is ok.
File -> Save Video -> If you use VirtualDub2, you can save in MP4 or MKV, otherwise AVI is the only choice.

Now upload on Youtube and share

[Alwyn]

Both use the same Elgato stick, but the "Elgato_Raw.mp4" is captured by Elgato software, while the "VirtualDub_Huffyuv.avi" is captured by VirtualDub? So, the Elgato stick can send unmutilated interlaced video, that is good.

That's not the way I read it. It reads to me that they've captured with Elgato/Virtual Dub/HUFF, then they've encoded the AVI into MP4 with Handbrake.

[Sharc]

a) confirm my huffyuv capture looks OK

Yes, your huffyuv capture looks basically ok to me (not knowing the quality of the VHS source and your VCR).
Interlaced TFF, YUV 4:2:2, 720x480, 29.97fps, Levels within legal range. So technically ok.

However:
- Colors (saturation) difficult to judge. Low light scene, proc-amp settings off the kilter.....?
- Did you have some noise filter or other "picture enhancer" switched on during capturing (causing heavy tearing/ghosting) in your VCR or passthrough device? If so, it's usually better to disable it. What exactly is your setup?

The rest can be done in post processing and encoding. I can't help you much with handbrake as I hardly ever use it.

[jagabo]

The field structure of the huffyuv file is intact. But it suffers from terrible ghosting (a poor noise reduction filter was applied somewhere along the line) and DCT ringing artifacts (it was compressed with a lossy codec before it became a lossless avi). It also suffers from dropped/duplicate frames, bad color, levels, etc.

QTGMC (AviSynth) deinterlaced, and x264 encoded version attached.

[Xhumeka]

Thanks everyone - really appreciate the feedback. To answer a couple questions:

Both use the same Elgato stick, but the "Elgato_Raw.mp4" is captured by Elgato software, while the "VirtualDub_Huffyuv.avi" is captured by VirtualDub? So, the Elgato stick can send unmutilated interlaced video, that is good.

That is correct - elgato_raw.mp4 is the file captured by Elgato's software, and virtualdub_huuffyuv.avi is captured by VirtualDub (using the Elgato as the capture device). As for whether or not the Elgato can send "unmutilated video", I'm not sure about that and the following comment leads me to think maybe NO:

The field structure of the huffyuv file is intact. But it suffers from terrible ghosting (a poor noise reduction filter was applied somewhere along the line) and DCT ringing artifacts (it was compressed with a lossy codec before it became a lossless avi). It also suffers from dropped/duplicate frames, bad color, levels, etc.

I had enabled noise reduction in VirtualDub as per this section of the guide:

Go to Video, and then Noise Reduction to enable noise reduction (OPTIONAL). This setting can remove noise on the video capture, but it will increase system CPU usage, and may result in dropped frames depending on the compression.

Virtualdub was not reporting any dropped frames, so I thought I could leave this enabled. Should I try disabling this and recapturing?

The comment "it was compressed with a lossy codec before it became a lossless avi" alarms me - nothing in my workflow should have introduced a lossy codec, so does that mean the Elgato device is not a good capture device for this task?

As for the hardware used to capture, the Elgato_raw.mp4 file was captured with a JVC HR-S7500U with TBC enabled and picture control set to "Auto". The huffyuv file was captured using a Panasonic AG-1980p. Both methods had no passthrough device, it was just VCR straight to elgato capture device (via s-video). I've ran capture tests in the past with both these VCRs and I could tell no discernable difference in quality, so I assumed it wouldn't matter if I used a different VCR for the huffy capture, but maybe I'll switch back to the JVC for huffy capture for a better apples-to-apples comparison.

Here are my virtualdub settings: https://imgur.com/a/Xs3MnON

And finally,

"What do you want to do with it? Stash for better times? Edit in your favorite NLE? Save to a thumb drive and give to your parents? Upload to Youtube?"

My goal is to "stash for better times" - i wanted to capture the best quality (that my hardware allows) possible and save to a USB drive for working with later. I want to make use of my VCRs while they still work well and worry about rendering to .mp4 and uploading to youtube down the road. Thanks again all!

[jagabo]

Looking again it may be that what I initially saw as DCT ringing may just have been a part of the ghosting from the noise reduction. Capture again without the noise reduction. That may take care of the ghosting and dropped/duplicate frames.

[Bwaak]

To me, noise and ghosting in the AVI were caused by high gain and low shutter speed of a camcorder struggling in low light, while Elgato software was doing cleanup and denoising, it this case it would be doing pretty good job.

@Xhumeka, do you have better quality sample shot in good light?

Also, does Elgato software have an option to output in 60p?

[Xhumeka]

Thanks again guys - I'll recapture using the recommended changes and will report back. Unfortunately I'm travelling for work for the next 2 days, so it will be later this weekend when I have a chance to recapture.

In the meantime, I have clipped another segment (Elgato_Raw_2.mp4 and VirtualDub_Huffyuv_2.avi) that is from the same tape, but better lighting (filmed outdoors) - let me know your thoughts, and thanks so much!

https://drive.google.com/drive/folders/1ylSWTdVsAXIMbYpkIkBwLwwKdwfPicy1?usp=sharing

[Xhumeka]

Also, does Elgato software have an option to output in 60p?

No, the Elgato settings are very limited - all I can change is the resolution ( 640×480 (4:3) or 640×360 (16:9) )

[jagabo]

The first mp4 file appears to have been blend deinterlaced. That loses half the motion and vertical resolution and makes everything look like a double exposure when things are moving. The second one has been deinterlaced but not with a blend. Still severe ghosting in the new AVI file.

[Sharc]

The new avi capture from post#9 has blown out/clipped whites (masking details, like from trees in the bright sky) and elevated blacks (greyish). The luma should be in the 16....235 range.
There are still ghosts and poor colors. Disable any filters from the chain when you re-capture, and keep an eye on the levels.


[Attachment 67780 - Click to enlarge]

[Bwaak]

Some weird stuff is happening here. Chroma shift on the left edge of the frame is encoded as vertical bands, is this Huffyuv error? Chroma combing makes it look as if it was shot with a tube-based camcorder, so chroma is one field behind. And there is constant ghosting in motion scenes. Either Elgato software is very good at fixing such issues, or there is something very wrong with the Huffyuv capture.

[Sharc]

Either Elgato software is very good at fixing such issues, or there is something very wrong with the Huffyuv capture.

The screenshot above has DCT compression artifacts (16x16 blocks). How was it taken? From which source?

The Elgato software did not really fix anything. It just deinterlaced, hard-clipped the luma at Y=235 and blurred (denoised) the picture further and encoded/compressed it as AVC, masking some of the defects which became apparent with the interlaced Huffyuv capture.
I think something is odd with the VCR or with its settings, or maybe it's baked into the tape (a dupe?).

[lollo]

The Elgato software did not really fix anything.

Yes, the Elgato software should not be used at all.

Back to the original question, capture lossless HuffYUV with VirtualDub without any processing/filtering or with AmarecTV.

In addition, avoid a GUI for post processing on captured files, but rather use AviSynth script, where you can have full control (Hybrid may be the exception for basic filtering as starting point)

[Xhumeka]

Thanks again guys - when I return this weekend I'll recapture a couple small clips using Virtualdub/Huffyuv with no filters on each VCR (HR-S7500U and AG-1980P) and see how it goes. Maybe I should also order one of the recommended USB capture devices so I can see if the Elgato is part of the problem. Which is the "best" available USB capture device for VirtualDub these days? Black Friday weekend, it's a good time to buy at least

[lollo]

ATI USB 600, Pinnacle 710 USB (no black friday here)
IOData GV-USB, Hauppauge USB-Live 2 (may be some discount here).
Most important element is VCR and then the condition of the tapes.

[Bwaak]

The screenshot above ^ is from the file @Xhumeka provided, deinterlaced with Yadif and upscaled to 960x720, but the source has the same artefacts.

To isolate the issue I would use identical hardware for both workflows, and since the Elgato version looks much cleaner, I would use the JVC instead of the ancient AG-1980. I don't think that Elgato stick is a source of the problem.

Macroblocking seems to be caused by Huffyuv, a defect I haven't seen since early implementations of DV, which is understandable as Huffyuv is about as prehistoric as DV. I would use a newer and better intermediate codec.

There is nothing wrong with GUI tools per se, they just provide UI.

It is common for consumer camcorders to blow out highlights, and digital camcorders routinely shoot in 16-255 range.

[Sharc]

Macroblocking seems to be caused by Huffyuv, a defect I haven't seen since early implementations of DV, which is understandable as Huffyuv is about as prehistoric as DV.

Huffyuv codec producing macroblocking? Really? That's new to me......

[jagabo]

Macroblocking seems to be caused by Huffyuv

Huffyuv cannot cause macroblock artifacts, DCT ringing, blurring, more or less noise, etc. It's truly lossless with YUY2 in, YUY2 out. I've tested it many many with different sources. Every time I've seen someone claim it wasn't lossless they were unknowingly causing problems themselves, usually with an unbeknownst to them colorspace conversion in their editor (for example YUY2 to RGB then back to YUY2 in VirtualDub).

The shot above isn't showing marcoblock artifacts. It shows posterization and gross errors in the video handling. I believe it's the noise reduction filter causing it.

[jagabo]

Back to the original question, capture lossless HuffYUV with VirtualDub without any processing/filtering or with AmarecTV.

This is important. You want the original data coming out of the capture device. Any filtering can be done later. You also want to disable any noise reduction in your VCR and any processing device in the capture chain. The only processing you want from the VCR is time base correction (when available) -- that cannot be done after capture. Everything else can be done much better with software processing after a successful capture (and you can't do it while capturing because it's too slow, you will drop frames).

[Bwaak]

Fair enough. If it is not Huffyuv, then it is something else, but two major issues with this video are undeniable: macroblocking/DCT/banding on the left, where there is severe pink and green coloration, and posterization. These issues are in the file provided by @Xhumeka.

[lollo]

Huffyuv codec producing macroblocking? Really? That's new to me......

He's joking

[jagabo]

Fair enough. If it is not Huffyuv, then it is something else, but two major issues with this video are undeniable: macroblocking/DCT/banding on the left, where there is severe pink and green coloration, and posterization. These issues are in the file provided by @Xhumeka.

It's easier to see what's going on if you stack the fields rather than interleave them:


[Attachment 67811 - Click to enlarge]

LWLibavVideoSource("VirtualDub_Huffyuv_2.avi") 
AssumeTFF()
SeparateFields()
StackVertical(SelectEven(), SelectOdd())

In any case, everyone agrees the huffyuv caps have many problems. We'll have to wait for new ones.

[Xhumeka]

Thanks again guys - when I return this weekend I'll recapture a couple small clips using Virtualdub/Huffyuv with no filters on each VCR (HR-S7500U and AG-1980P) and see how it goes.

Alright you video wizards, here are the clips I promised... These are the same clips as before, the only difference is the "noise reduction" filter in VirtualDub has been disabled. There should be no filters applied to these clips:

AG-1980P Clip 1
AG-1980P Clip 2

HR-S7500U Clip 1
HR-S7500U Clip 2

Link to all clips: https://drive.google.com/drive/folders/1ylSWTdVsAXIMbYpkIkBwLwwKdwfPicy1?usp=share_link

Let me know what you think - thanks again all!

[Sharc]

After a first glance (clip2): Much better now. The ghosts/tearing has disappeared.

I would prefer the HRS75000U because the video levels are legal even when converted to RGB. It also has less halos than the AG1980.

The AG1980 would need some levels tweaks to prevent clipping and illegal RGB. Can be done for capturing possibly via proc-amp settings or in post processing, e.g. reduce the brightness and contrast a little (still some remaining RGB crushing though).

For post processing I would suggest to apply some chroma sharpening and spatial chroma shift for both to fit the luma. Color tweaks (hue and and saturation) to taste.

[jagabo]

I agree with Sharc, the HR caps are better. I didn't notice any dropped/duplicate frames. And with all the caps the problems caused by the noise reduction filter are gone.

[lollo]

I agree with Sharc and jagabo. Capture lossless and without filter as recommended

[Xhumeka]

Amazing, thanks guys! Now I can continue re-capturing important content with huffyuv lossless codec knowing I'm not wasting my time/effort, and can rest assured that I am preserving in the best way possible (with my current hardware).

Thanks once again all!

[Bwaak]

It is interesting how on the HRS75000U capture fields have different color temperature, while on the AG1980 they look the same. I though that artefacts like this would be caused by a camera sensor, not by a playback device.