Recording, encoding processes for demanding client?
 

Hello DigitalFAQ.

A few months back, I made this thread about capturing 24-bit audio in VDub (which I figured I could just record in 24-bit in parralel): https://www.digitalfaq.com/forum/vid...ng-24-bit.html

But, now, my problem focuses onto something else: interlacing, PAR, Adobe Premiere and client's demanding preservation format.

Now, the client has finally sent us their tapes to be digitized. Their "preservation" standard is extremely demanding to the point of "WHY": AVI v210 10-bit 4:2:2 interlaced.

They even go out of their way to say "ProRes and MXF/JPG2000 not selected because they are more production industry standards."

Premiere supports interlacing. This is not the issue. The issue is that, with Premiere, with Huffyuv, it will randomly insert garbage frames. If you look at the footage with, say, PotPlayer or VLC, it plays back with no such frames. ON TOP OF THAT, it takes an ETERNITY to add the files to a project, especially if the footage is very long. Not to mention scrolling and editing... since I have to sync the new 24-bit audio to replace the vdub 16-bit ones.

As I was writing this, I checked the option to force RGB in huffyuv and tested with our 13min long test tape. Premiere gobbled it up and didn't show any garbage. PotPlayer shows it was interlaced (PotPlayer uses the "Tree" setting of MediaInfo). However, Premiere sees the footage as Progressive. Fine. I'll make it interpret it as interlaced UFF and correcting the PAR to 0.9091 to make 4:3 instead of 3:2.

Then, I setup the encoding like they want (see screenshot 1 and 2).

But then... opening the resulting file... it's still 3:2 despite setting 0.9091. The combing effect becomes much more pronounced versus huffyuv (screenshot 3 = huffyuv, screenshot 4 = client settings)

So now... I need help figuring out how the ... am I going to make this v210 file while making it look as good as the initial huffyuv while being able to come in and replace the 16-bit audio file for the 24-bit one recorded in Audition while the tape is recording. I'm open to all software as long as it has a GUI. They need a stupidly high bitrate de-interlaced MP4 video with it (for "access") which proper de-interlacing help would be great help. Please note that this client will not be changing their mind."

Capture Card: ATI TV Wonder 600.
Computer: Windows 7 HP x64, Core i7 950, 12GB DDR3 RAM, GeForce GTX 680, SATA OS drive (90GB SSD), writing to USB SSD through USB 3.

Please note that I may not be answering on weekends as I see my postings here part of my job and figuring things out.

Thank you.

As soon as I read the title, my first thought was "fire the client". :laugh:
I've not yet read the thread.

Replying as I read...

- Interlacing is not really a choice. It's either interlaced, or deinterlaced/progressive. Deinterlacing is either QTGMC or worse, with some processing options and constraints.
- PAR is PAR.
- Premiere will be Premiere.
- "client's demanding preservation format" usually means he/she knows nothing, and blindly trusts Google/ChatGPT or other random non-sources

The more I read, the more "fire the client" vibes.

Premiere itself doesn't insert garbage frames. It's relying on the codec install. Are you using the actual Huffyuv? Not the fake/messy FFMPEG version, not MT, not 64-bit, not CCE hacked, etc. The original Huffyuv by Ben R-G. (Lagarith is the codec known for random bad frames, not Huffyuv. Odd.)

There's always alternatives.

If I were using a PITA NLE (Premiere or otherwise), I'd just extract everything to uncompressed. Compared to the 2000s, even the 2010s, we're living in an era of infinite CPU and drive space, in the context of SD video processing.

V210 is going to be a problem. From a quick Google search, even LoC doesn't use it (or want it).

When I have somebody approach me for work, asking for something wrong/weird/stupid, I ask where they got that information, because their request is "odd and non-standard". In 99% of those cases, it's Google/ChatGPT/randos. I tell them what is available, and ask what they want. Most proceed. Some stubbornly stick to the request -- and also always argue it, as if they can make me do what they want. I have no time for their BS. I wish them well, end the conversation, delete, move on.

These people will continue to seek their Moby Dick, until they find a service willing to do it. It never ends well. Many of them crawl back to those who denied the initial requests, wanting to get it re-done as was suggested from the beginning. They caught Moby Dick (or the car bumper), and had no idea what to do with it, and were surprised it didn't look (taste) as expected.

Unless you're getting like $10K+ from this project, fire him. (Honestly, even $10K may be low.) Writing "this client will not be changing their mind" is just wishful thinking. I'd bet money on it. It gets easier to 'fire clients" the more you do it. You start to learn to spot them before even accepting projects, and just decline their work altogether. They can be somebody else's problem and stresser.

What gets me here is that, while the ATI 600 USB is an excellent card, it's not the "end all, be all" of ultimate quality. It captures legalized, and the values are imperfect. It's no AIW, nor a broadcast appliance. Some of the demands made here exceed what the card did, so it's just insane.

Going as I go as well...

* I know interlaced is not really a choice. They want to keep the interlacing for an interlaced format which is understandable.

* PAR is PAR... well. You know... I need to set the correct one in post-processing. A VHS is 4:3, not 3:2.

* I swore I took a screenshot somewhere of that glitched from when I used the same setup to record some Umatics for a MUCH less demanding client. It turns all green. You see the objects and all but like ugh. If it happens again, I'll take a screenshot and show you.

* Yes, I know I installed that Huffyuv from you. Upon checking the checksums, it seems I have installed the 64-bit version on my current Windows 11 machine (R9 5900X, 32GB DDR4, Quadro RTX A4000). I do not remember what made me install that one instead of the old 32-bit non multithreaded one. I know Premiere works with them like this.

* In the documentation that was given to us for video digitalization, the last modification date is "June 17, 2021" and that's written in text in the MSWord file and not the file's "creation date". So, before ChatGPT really became a thing and I imagine that this was decided way before then.

* What NLE would you choose if you were stuck in my shoes?

* I'm very aware that no other preservation bodies even consider V210. It's either .mov (codec used is not even mentionned. Could be ProRes, H264, H265... who knows?) or MXF.

* This client is a lifeline for my salary. Their entire library must be already in those formats (and probably fake 24-bit audio, upscale from a 4:2:0 video and they probably don't realize). Also, I'm not the one that seeks jobs. I'm just the guy that gets told to digitalize and what settings to use if the client has particuliar ones.

Yup. I realized that as well. The card does 4:2:2 (YUY2) but at 8-bit and most videos on the internet today are just that... 8-bit 4:2:0.

Digitize as YUY2 AVI 8 bit (maximum quality possible in my opinion for VHS) and then convert to whatever in Adobe :)
10bit for VHS? Very demanding client! Did he ask to capture VHS at 4K too? ;)
Only, after importing that in Premiere, do not forget to set "interpret footage" field order "upper field first" for PAL and "lower field first" for NTSC for imported YUY2, because otherwise Premiere will interpret it as progressive, because no field marks in this format.

It depends. If captured standard Full D1 720x480, then it has 16 extra pixels that needs to be cropped before 4x3 resize/processing. Some cards allow 704x480, but then it may not crop right either. Resizing can be messy if done wrong. That matters far more than bit depths and codecs.

"all green" is an internal error, generally inside the NLE in this scenario.

It's not really a "64-bit version", but an entirely different FourCC codec. It's Huffyuv in name only. In order to make the actual Huffyuv install the 32-bit/x86 for 64-bit/x64 usage, the "easy button" is to use the hofmand installer.

I wasn't really focusing on ChatGPT itself, but really just random "research" (term used loosely) using Google/whatever. The person didn't know (and still don't know what they don't know), and just blindly trusts random information as good and accurate. That takes a certain type of personality.

What I've often had to do is:
- capture Huffyuv, retain as master ingest files
- VirtualDub2 convert to the unofficial ProRes422
- transfer those files to Mac, edit in Final Cut

Sometimes that's the path of least resistance. And my own M2 Pro is more powerful at editing than my Windows systems (where I no longer have Adobe subs, only final pre-sub CS6 versions).

- I hate MXF, but it is a standard, just not a very good one. Both it and DNxHD files are obnoxious.
- It's really 50/50 on ".mov" being uncompressed or ProRes -- but not sure which ProRes. And it does leave open the door for compressed, including the awful Apple version of H.264.

Hmmm....
I need to know more.
Are you doing the Premiere work, then exporting? So they'll never see the Premiere step?
So you're ingesting Huffyuv, moving to Premiere (editing?), exporting to crazy specs?
So they never get the Huffyuv, only the final Premiere export files?

I think this can be done by "adding" a step, but ultimately saving you time and grief. :devil:

VHS source is not even 8-bit, more like 6-bit dithered.

Think of VHS as a cake, 6-bit.
Capturing to 8-bit adds 33% icing to the cake.
10-bit adds 25% more.
12-bit adds 20% more ... and the "cake" is now literally half icing.
That cake sucks. :laugh:

I think we should converted to whatever for Adobe. Then export to whatever as output.

That's not correct. Only DV is BFF ("lower", bottom), everything else is TFF (top, "upper")

Sorry, my mistake, yes upper for both of course.

SHGMC_2, I do such capture.
So, send me this customer. Well, mostly his money.[:D]

I think they'd be hard pressed "not to be ok" with ProRes422HQ which does preserve interlacing. It is generally visually lossless and is 10 bit, unless you are running through an equipment bottleneck such as an 8 bit TBC or other ADC that is less than 10 bit.

4:2:2 does have an advantage in the raw capture state as long as you upscale prior to making it the 4:2:0 delivery format. 4:2:0 basically means you don't have any color information recorded for every other horizontal line. If you were to say upscale 480p/4:2:2 to 960P (or greater)/4:2:0, you haven't actually lost any color information from the original because the horizontal lines are doubled before then getting rid of the color information on every other horizontal line.

If it was me, I'd start with a ProRes422HQ capture, crop, and upscale to 1440x1080 progressive 59.94fps (essentially a bob QTGMC deinterlace) at 15000Kb/s. If the client wants the raw capture, they're going to find ProRes to be much more generally compatible with all nonlinear editors and operating systems. With the upscale, you don't have to worry about pixel aspect ratio anymore either since it'll be square after that. There's also a way to just use FFMPEG to set the PAR flag without a re-encode, but for that to be helpful the player or editor would then need to pay attention to that flag (which not all of them do).

That's not possible here.

Why? That's not the job specs.

I guess in the end you could convert from ProRes422HQ to whatever format they want since it is visually lossless, 10 bit, and interlaced. Could ten set the DAR to 4:3 using FFMPEG assuming that format they want is supported by FFMPEG as a final step.

If there's an issue as to what the characteristics of this "V210" thing is, I'd have them send you a short clip of something they already have that does meet their requirements so you can figure out what the codec and other specifics are using mediainfo.

Back from weekend!

(going as I read)

* I see. I mean, In Adobe, you can make a 704x480 canvas and have the video aligned in the middle or either side depending after having the footage interepreted as 4:3.

* I figured as much because even converting straight from Media Encoder would render garbage "green" frames. Now, using the "RGB" mode in huffyuv, it just insert a previous frame. Meh.

* This one: https://github.com/hofmand/video-codec-installers ? If yes, do I need to install the previous one before? If yes, how do I do so? Did you explain somewhere?

* Yeah, well, I certainly didn't decide and my superiors definitely won't try to argue with them. :\

* Interesting. Might keep ProRes in mind for being compatible with interlacing. I'll try to keep a lossless chain for this client... even if it does not make sense. If there's a huge issue, I'll talk with my boss about it to see if I could take shortcuts or something.

* Awful Apple H264? I dunno but Premiere, I'm using Voukoder which allows me to use NVENC H264 with .mov. Are you talking about Apple's x264 being awful? Because I definitely don't have trouble believing you in this case.

Pretty much, yes. They don't see the original capture file. I replace the original recorded audio with the 24-bit audio I recorded in parralel. Crop the beginning and the end. Then export. That's the plan for them anyway.

For other clients, I'd be using the AIW setup we have. Get the MPG file, put that in premiere, Right-click clip -> Fields Processing -> Always Deinterlace, Raise the volume a bit. Depending on the job, I might crop the overscan and empty sides then center the remaining picture (keeping a standard resolution). Then, export to MOV NVENC H264 / PCM (VBR Bitrate - min. 2200kbps, max 3250). For 99.9% of people, that's PLENTY. Most will be happy with having a picture at all.

But not this client...

* Pfffff. 6-bit. Yeah, I agree. VHS were always bad quality to me. Seeing 6-bit does not surprise me.

* Convert for Adobe... yeah, definitely. It has caused me so much headaches.

Upscaling with the ATI TV Wonder 600? lolno

I believe they have no idea what is 24bit audio and what for.

No one can hear difference between 16bit and 24bit, but I attached Swiss Sound from 1994 where 24bit is explained. Yes, I always digitize analog audio at 24bit (32bit float) and 48khz (broadcast wave standard). If someone want 192khz - OK, no problems, we can! This is why dogs lick their balls - because they CAN! ;)

Better convert sample type from 16 bit to 24. You can go out of sync with your method.

P.S. I mean no one can hear difference between 16bit and 24bit in final release. But 24bit is very important as initial format. If you will not edit audio track, it is really do not matter, just convert and forget.

P.P.S. However keep in mind your client can always ask to explain your recording technology, so in this case you should have something to show how you do it (quality do not matter, because they know nothing). I would avoid such clients, because I avoid lying. I would say - my technology is following and "take it or leave it".

1. Yeah that's the right github page for the modern method of installing huffyuv, just uninstall the previous huffyuv codec and follow the instructions listed on the github page.

2. Think Aramkolt means upscaling after you're done capturing your raw file, ie using a program like Selur's Hybrid to do upscaling, deinterlacing (plus any other restoration using avisynth/vapoursynth) and encoding all in one program.

I know why they want 24-bit audio. They want better resolution fidelity of the analog audio being converted from a pure waveform to a digital "stairs waveform" - you know more steps = more details. It's not a matter if you can hear the difference or not. However, the quality of the analog audio on the VHS tape is what determines, for me, that VHS is only really worth 16-bit.

As for the desync... I think it's only the case for the first 2-3 seconds and after that, it's pretty much sync'ed. Checked the waveforms of both files through the 1h video file and, besides the beginning, it keeps lining up. If I had seen that it wasn't sync'ed throughout the file, I wouldn't have used the parralel recorded audio file. Please note that VDub takes the audio from the same audio source that the separate audio feed takes its audio from. Which is not the capture card itself because 1. Capture Card has no volume input control and 2. Capture Card isn't seen in audio software as an input. A dedicated sound card is used.

Like I said in the original thread, I'm not one to half-_ss and do fake encodes. I don't want to fake a 24-bit audio file.

I can't "uninstall" it as I used the registerdll method via command-line LS mentionned here: https://www.digitalfaq.com/forum/vid...l-huffyuv.html I wouldn't know how to revert that tbh.

I tried Hybrid in the past and I found it to be a clusterf**k to use. Settings wouldn't work 3 out of 4 times even if most settings were left on defaults. And, now, even after re-install, the program refused to launch. So, I uninstalled it.

Also, I'm allergic to CLI. I had to ask """""""""""AI"""""""""""" to make me FFMPEG command lines. So, Avisynth is out the window (I couldn't tell but is Vapoursynth a GUI version of Avisynth?). And, really, they do ask for an "access" MP4 file (16 mbit bitrate for 480 lines video...) which the deinterlacing in premiere did the job. They didn't even mention what audio they wanted for that. So, I just threw AAC 320kbps in there.

Really, my big issue is to get the right encoding out of Premiere because, even if it's all set to NTSC 0.9091 pixel ratio, UFF (including the clips themselves), the resulting AVI file is still read as 3:2 by Pot Player despite me going out of my way to make sure all settings are correct for the final output. Uugh... I hate these settings so much... never had such issues before because all other clients we ever got for VHS never really asked anything in that matter. Ugh. And the interlacing look even worse after being exported with premiere in the v210 file.

I know it's a bit late, but I thought of a workflow that could give you 24-bit sync'd with your capture. Some ATI AIW cards send their audio out through the sound card, and you could use something like a Tascam US-122L or US-144 in XP to capture in 24 bit.

As someone who has worked in video/audio and had to sync 3+ hour long audio recordings with multiple cameras, I've come to deeply dislike anything that isn't timecoded. So if you can build an XP rig in a day and it's a big client, it might be worth considering. Sometimes you think it's in sync but it's really off by some ms that you don't quite perceive right away. It's so mind screwy sometimes I would rather just avoid it if at all possible.

More details for what? :) Read that 1994 Swiss Sound, there are described benefits of 24bit in digital mastering process. Before conversion to final 16bit for CD mastering, for example. Yes, now 24 bit is not a problem, so 24bit is everywhere. There is nothing wrong with 24bit, I digitize even 78rpm from 30s in 24bit, my Universal Audio 2192 ADC's support only 24 bit, but - you (and your client) should not bother about 16bit. Ask him, it is 16bit 100db SNR what bother him and why he needs 140db? From VHS :laugh:

I guess you can go ahead and try the installer anyway, might override those files. Hope it'll work

Manually search for "huffyuv.dll" on the system, and delete those. These files are in Windows system folders, so the OS "security" may fight you some.

Once done, use the hofmand installer.

Or get Blackmagic intensity (all them support 24bit) to explain to client how you do it, but at the same time, quietly... ;)

I can send you one Intensity shuttle. Nightmare, but supports 24 bit (if I remember right). You could show that to your client (and demonstrate, it is working with terrible audio delay, but for demonstration purposes does not matter). After your project you could send it back to me or throw out, no matter. Shipping cost on your account ;)

This method is too much manipulation. It doesn't crop 8, then resize pixels to 4x3. The Premiere way resize pixels, then stretches pixels on crop. This is why pre-NLE work is best done in VirtualDub or Avisynth+VirtualDub. Are you actually "editing" the captures, or just "scissors snipping" the ends of the capture? If snipping, Premiere is 100% the wrong tool here.

RGB is making matters worse, don't do that. However, Premiere is not native YUV, colorspace conversion is already happening in the background.

I'm my own superior, and sometimes I argue with myself. :laugh:

Apple's Quicktime H.264 is awful, notoriously so. (BTW, x264 is not H.264. Not interchangeable terms. H.264 is licensed, x264 is unofficial unlicensed. --- And you may need to be careful with your client. Or rather, your client needs to be careful. MPEG-LA licensing should be reviewed by you and them.

Are you visually aligning external audio with Premiere? Is that the only reason for Premiere? We may nede to massage your software workflow. On the surface, your method may be "easier", but another "more complex" method may actually be long-run easier. (Been there, done that. I go for path of least resistance, even if a few extra steps.)

Your superiors aren't do the work. You are. You still need a blunt conversation with them after this project is done, and "no more" needs to be the outcome. You've been screwing around with this for months, because the requests doesn't match the source. It's non-standard, to the extreme that other large orgs do not use it (on purpose). I've worked at large orgs. I've never been afraid to say "no". I don't live life in fear. If anything, that's landed more respect and pay raises.