VideoHelp Forum
+ Reply to Thread
Results 1 to 13 of 13
Thread
  1. Hello!

    2 parts to this post. First is a question about proc amps. Second is a question about how a digitizing workflow could be optimized for realtime processing, specifically live streaming.

    Background: I first started looking into digitizing analog signals for the purposes of retro game capture/streaming (this is still a big focus, but no longer exclusively so - I have started playing with camcorders too, and basically want to build out my knowledge and apply it to a variety of scenarios. Definitely in the market for a VCR as well, family has some old VHS tapes I want to try my hand at preserving).

    There is a ton of misinformation about the nature of analog signals from the video game community, and I've been slowly unlearning the garbage. But there are certain questions I have that are - I think - specific to game capture for the purpose of live streaming gameplay.

    QUESTION 1

    Stepping back, though; question 1 is just about proc amps.

    If I want to digitize some footage from a non-interactive video source, this is the setup I would use:

    Video output in either composite or s-video -> **PROC AMP QUESTION GOES HERE, SEE BELOW**-> Hauppauge HVR 1250 capture card -> record lossless interlaced signal via AmarecTV or VirtualDub (have been experimenting with both) -> any upscaling, deinterlacing, editing, whatever can happen in post so long as the recording is good.

    (I'm not worried about audio at all here btw, I don't even use the Hauppauge for that... I have a scarlett focusrite 4i4 for music stuff, so I just capture stereo audio using my interface and have had good results with that. I'm solely focused on video atm.)

    **Question: I've found the Hauppauge HVR 1250 proc amp settings appear really effective, particularly the brightness and contrast. This is in complete opposition to, say, the GV-USB2 which I was using before which is pretty clearly just doing some mediocre filtering in post. Is the Hauppauge card doing something with the signal pre-digitizing, or does it just have better processing on the digital side that makes the proc amp settings appear more effective? Does this have to do with the fact the card has 3D comb filtering? I'm still learning how this stuff works and it's hard to find good information among all the garbage.

    My reason for asking this is I have been toying with whether to get a proper analog proc amp but if the Hauppauge is already being this effective it might be a waste of money. Have no experience using a real one. Unless the Hauppauge card is, effectively, a "real" one and that's why results look so much better? I don't even really know what I'm asking here, other than that I want to understand what my card is doing, and what a separate hardware analog proc amp might potentially do differently.

    The good proc amp settings really help for live stream purposes since you don't have the option to edit in post... but for just recording footage, I want to better understand what kind of processing is appropriate to happen where in the signal flow.

    That concludes question 1.

    QUESTION(S) 2

    Question 2 is actually a lot of questions about the fact that, for live streaming, obviously the workflow can't be quite the same. I know that's not necessarily the focus of this forum, but I cannot for the life of me find knowledgeable sources about this stuff from the video game community so I'm hoping it's ok to ask.

    For example, I want to stream GameCube game footage being output from my Wii. My current setup is thus:

    Wii S-Video -> Radioshack signal amp/splitter -> CRT TV
    Radioshack signal amp/splitter -> Hauppauge HVR 1250 -> [SOFTWARE STUFF I WANT TO OPTIMIZE GOES HERE] -> the internet

    My questions are about optimizing that software stuff for the best trade-off between quality and the fact that realtime processing is necessary for streaming.

    [Bonus question... The splitter/amp I use is one that I found recommended by Melee streamers and it seems to work really well, but if I go further down the rabbit hole I'd be kind of curious to test other ones... anyone have experience with any?]

    Actual questions, in no particular order:

    * Scaling. The source video I am capturing is, obviously, 480i. I can get it to approximate 480p from deinterlacing [see next question]. But most people viewing a stream are going to be doing so on 720, 1080, etc. monitors. Is it better to scale the image on my end and stream at 720 or 1080, or should I stream source and have the scaling happen on the viewer end? There is something in my brain that says "no, do not stream in 480" but I don't really know why and I'd like to better understand - is it more likely that I can do software scaling on my end in realtime that will be more likely to look good in higher definition? Or will most viewers' screens do a good enough job scaling that I may as well stream in 480 and call it a day?

    From pure testing I have been experimenting with using OBS's bicubic scale filtering and resizing the image. Are there better ways to resize?
    Or, heck, would it be a better idea to just use a 1080 canvas and have the footage be exactly doubled, and use the extra space to put a border around it?
    Final complication on this point: it's preferable to squish the footage to true 4:3 so it looks right. I found that applying a 4:3 filter in OBS caused it to stretch vertically instead of squish horizontally?? How can I make sure that the footage is being processed as little as possible through all this? Processing is obviously going to HAPPEN; I just want to make sure I understand exactly where and why it is happening.

    * Deinterlacing. I've experimented with two workflows: one is feeding the card into AmarecTV, using AmarecTV deinterlacing (diferent settings for different games), and window capturing that in OBS. The other is just feeding the capture card signal directly into OBS and using OBS's deinterlacing. These both have pros and cons, but I was wondering if there are any better options for near realtime deinterlacing (have been using Amarec's "role-playing game" setting most frequently, and OBS's Yadif2x). Basically open for recs, I know this forum has a lot of recs for high quality deinterlacing in POST, but wanted to just pick your brains in case any of those solutions could be potentially used for live streaming as well.

    Last question:
    * Source video. Somewhat related to the above but want a specific question about it... I know if I were recording Amarec is going to give me more accurate footage by a landslide, but for live streaming, is the preview Amarec gives actually significantly better than just using OBS directly? Maybe its 4:3 resize and/or 200% window size would be better than OBS resize and that's a point in its favor? Now, I have done some messing around and I'm perfectly capable of just going "well I think I like this the best" and making subjective decisions. My purpose in asking is whether there's anything going on behind the scenes I might like to know about that makes one a potentially better solution for reasons I haven't thought of.

    I considered trying out VirtualDub but probably because of Win10 compatibility issues I can't get the preview on it to work at all. It records beautifully but the preview just does not work. So pretty out of the question for livestreaming atm.

    Sorry for the super long post. I reached a point where I felt the more I was learning about this stuff, the more I realized I don't know anything LOL.

    If the streaming questions are not suited for this forum that's ok btw I still want to know about the proc amp stuff. Also, I don't expect anyone to respond to every single question here, if anyone just has thoughts on any of the topics discussed I would love to hear them.

    Thank you!
    Quote Quote  
  2. It is better to scale to at least 720p because at this size YouTube turns on 60p. Although you haven't said that it is going to be YouTube, maybe you are using some other platform like Twitch with its own rules.

    IMO, Yadif is good enough and it works in realtime.

    I prefer watching 4:3 video as 4:3; if it has wrong proportions it is just a sloppy job. I don't know much about OBS, I was able to use successfully exactly once.
    Quote Quote  
  3. Member Cornucopia's Avatar
    Join Date
    Oct 2001
    Location
    Deep in the Heart of Texas
    Search PM
    Which one has less artifacts? 480p @ 1Mbps or 1080p @ 1Mbps? (all else being equal)
    Answer: 480. For that bitrate, more bits are applied to smaller rez, so quality per pixel is better.
    But obviously, if you can have 1080p @ much higher bitrate (e.g. 8Mbps), it might more than make up for the higher bit requirement and possibly also the quality hit from resize interpolation.

    So, much depends on how much bitrate you can consistently stream. And where this is going to end up (and also IF it's going to be recompressed, as was mentioned about YouTube).

    Good proc amps occur in the analog stage.
    It is possible to have a great digital proc amp, but that would require the capture device was preliminarily quantizing to linear float, or a very high integer bitdepth. And then the proc amp section would need to operate also in that same stratosphere, prior to downgrading to normal bit depths for saving. Otherwise, you can crush blacks and/or whites due to not enough remaing dynamic range to start with, or you can get banding due to the insufficient bitdepth to calculate changes, or both issues. With analog, you do have SOME limits (self noise, and amplification overload), but for most purposes, it is much more widely adjustable without adding artifacts.

    I will say once again that OBS is not optimal for working with analog sourced video, this has been gone over many times. However, I am a realist and know that certain applications and workflows sometimes require a change from the normal optimal way, and certainly OBS has a place in streaming production.

    Personally, I am not a fan of resizing ahead of time just for resizing sake. If there is a purpose (YT quality cheat, editing window interop compatibility) that's another matter. But you should pick what works best for you. Just understand what you are compromising and when. Remember though that most end viewers have systems that can/will resize to ~equivalent full screen upon playback, and the difference is that then there is no additional generational loss penalty.


    Scott
    Quote Quote  
  4. Ooh these have already given me a lot to think about. I can stream to Twitch at 480p60 (afaik) but wasn't aware YouTube requires a resolution of at least 720 for 60fps.

    I just realized that the bitrate I've been using (~3200) is below the recommendation for 720p60 anyway -- now I've gotten better internet since then so maybe I can push it now, but regardless, I'm gonna keep settings the same just to test and do some 480p60 streams, then view fullscreen on a 1080 monitor and see how it looks in comparison. Makes complete sense that given a limiting bitrate, focus all that information on only the pixels I actually need. I'm probably going to have a completely different workflow for recording videos (as opposed to livestream) anyway so maybe YouTube's 720-and-up limitation won't matter.

    [and on the other hand, if I find my current Internet is able to give me a high enough bitrate for really nice 720p60 streams, I can see how that looks scaled on my 1080 monitor next to the 480.... best approach is to just do it, huh]

    Thanks a lot for the input, will continue reading responses here but also gonna do some more testing.

    (also thanks for the info on proc amps -- don't quite know yet how to verify what my card specifically is doing but what you've said in general makes sense. i'll probably get an analog proc amp eventually.)
    Quote Quote  
  5. Member
    Join Date
    May 2005
    Location
    Australia-PAL Land
    Search Comp PM
    I've found the Hauppauge HVR 1250 proc amp settings appear really effective, particularly the brightness and contrast. This is in complete opposition to, say, the GV-USB2 which I was using before which is pretty clearly just doing some mediocre filtering in post. Is the Hauppauge card doing something with the signal pre-digitizing, or does it just have better processing on the digital side that makes the proc amp settings appear more effective? Does this have to do with the fact the card has 3D comb filtering?
    I'm curious about your comments re the GV-USB2. I use this as my go-to stick for capturing VHS. Could you explain more your comment " is pretty clearly just doing some mediocre filtering in post"? I've found it controls brightness, contrast and saturation just the same as the other capture cards/sticks I use.
    Quote Quote  
  6. Originally Posted by Alwyn View Post
    I've found the Hauppauge HVR 1250 proc amp settings appear really effective, particularly the brightness and contrast. This is in complete opposition to, say, the GV-USB2 which I was using before which is pretty clearly just doing some mediocre filtering in post. Is the Hauppauge card doing something with the signal pre-digitizing, or does it just have better processing on the digital side that makes the proc amp settings appear more effective? Does this have to do with the fact the card has 3D comb filtering?
    I'm curious about your comments re the GV-USB2. I use this as my go-to stick for capturing VHS. Could you explain more your comment " is pretty clearly just doing some mediocre filtering in post"? I've found it controls brightness, contrast and saturation just the same as the other capture cards/sticks I use.
    I'm actually in the middle of making a video comparing the two (and my post here is motivated by a desire to make sure I know as much as possible on the subject before posting it) so I'll have a really clear demonstration to show you when I'm finished with it lol. The biggest difference is the contrast, GV-USB2's is pretty lackluster imo and have been blown away by how responsive the Hauppauge's was.

    But that's probably because Hauppauge HVR 1250 is an on-board card that I believe can take much better advantage of my computer's processing power.

    GV-USB2 is great for the record I used it for like 3 years with no problems. I'm just really loving the Hauppauge I just got because those granular controls are awesome for tweaking the visuals for game livestreaming. The tweaks I make for visual effect for that purpose are probably counterproductive for VHS archiving as they are more for visual "wow" and almost certainly cause clipping.
    Quote Quote  
  7. Member
    Join Date
    May 2005
    Location
    Australia-PAL Land
    Search Comp PM
    Thanks Arcana, I will view with interest. I normally set the levels with the histogram before capturing and then don't adjust them again.
    Quote Quote  
  8. UPDATE ON STREAMING SCALING:

    Hahahah oh wow I finally figured out the biggest point of confusion/frustration for me. I read an older thread on here about pixel aspect ratios, and therein lies my problem.

    I do not think OBS (and probably Twitch/other livestream platforms) support pixel aspect ratios that are not 1:1. Maybe I'm wrong. I would love to be wrong. But I cannot for the life of me find anything that implies it does.

    So my options are:

    1. Stream source (720x480) and deal with the fact it is stretched; trade-off is wrong aspect ratio but transmits source pixels exactly and better image for lower bandwidth.

    2. Stream higher resolution canvas and upscale/stretch 720x480 source on my end so 4:3 aspect ratio is correct, pixels will not be completely accurate but in this case I think higher resolution = better approximation. Decision between 720p or 1080p will depend on whether encoder/bitrate can handle higher resolution, versus how accurate the scaling looks for each.

    Sooo I actually did a ton of test streams with varying settings, just to see. I promise I am not doing this to shill my twitch. It's just the best way to test how a stream is gonna look is by doing it. These were all using the same bitrate, 6000Kbps, which is recommended for 1080p60 (but since my connection could handle it, no reason to lower for 720).

    1080p canvas tests:

    https://www.twitch.tv/videos/1504501655 = vertical doubled to 960, horizontal stretched to 1280
    https://www.twitch.tv/videos/1504503386 = horizontal doubled to 1440, vertical stretched to 1080

    720p canvas tests:

    https://www.twitch.tv/videos/1504502503 = horizontal preserved at 720, vertical stretched to 540
    https://www.twitch.tv/videos/1504504165 = vertical stretched to 720 to fill canvas, horizontal stretched to 960

    I'm gonna need to take a while looking at these on different screens of different native resolutions to see what looks overall best in the widest variety of circumstances. For example, how does the 1080 resolution look downscaled back to 720 if I'm viewing on a device that can only handle 720? Versus viewing 720 resolutioin upscaled on a 1080 monitor? There's no perfect solution my goal is just to overall find something that seems like the best fit for a wide variety of scenarios.

    [edit: tentatively, I think the 1080 canvas with the vertical pixel count doubled is possibly the best bet? an integer multiple of vertical resolution seems like a good idea because of the nature of interlaced footage. looks great at 1080p60, seems to downscale back to 720 acceptably well.]

    [edit edit: on the other hand, non-partners on Twitch.. aka most people... don't have a guarantee that viewers can select stream quality. so only streaming in 1080 and a higher bitrate can lock out viewers with a slower connection. So maybe it's ideal to optimize for 720 and a lower bitrate. but you can't double verticals at that resolution.. alas. Well, at least whatever I go with it'll be an educated decision. I'm almost tempted to take the L on a raw source 480x720 stream having the wrong aspect ratio as a compromise...]

    edit edit edit: screw it, did a test of 480x720 60fps stream with no scaling, only deinterlacing, may as well be thorough. 3000K bitrate for this one. Now I do think the lanczos upscaling before stream looks better than this, possibly just because it's a better algorithm than whatever my browser is doing, but here's a link for reference = https://www.twitch.tv/videos/1504591994

    Thanks again for all the input here. Will update if I realize anything else major, and Alwyn I'll let you know when the comparison video is done
    Last edited by ArcanaXIX; 15th Jun 2022 at 14:10.
    Quote Quote  
  9. Sorry to double post, I think I figured out why those digital proc amp settings look so much better on the Hauppauge as opposed to the GV-USB2. My computer unfortunately can NOT screen record while also displaying the video and wave form at once (using QCTools to look at some videos I recorded next to a wave form analyzer -- note these videos were recorded in OBS because they were for a video I am making on livestream settings, not optimized for offline recording), so I had to just record this with my phone. Brightness tweaks move the whole waveform up/down, contrast tweaks shrink/expand the range of the waveform.

    But, three big things I'm noticing:

    1. Hauppauge is a LOT more granular/precise. One tick left or right has pretty minute changes, whereas GV-USB2 one tick changes the wave form dramatically. So you can be a lot more precise and nitpicky about the setting.

    2. Hauppauge looks like it picks up more level distinction in the source signal to begin with - there are distinct lines in the wave form that are pretty separate, whereas GV-USB2 the majority of information is clumped together at a big blob at around the same level. Can't emphasize contrast when there isn't much contrast in the data to begin with.

    3. Algorithmically, this is the big kicker - it really looks to me like lowering contrast on the hauppauge acts more dramatically on the extreme highs/lows but leaves the mids relatively unaffected. This has the effect of bringing up or down the overall contrast but maintains the detail in the middle, which is where the vast majority of important visuals are (not going to be staring at extreme brights or darks for most of gameplay). The GV-USB2, on the other hand, looks like it just applies a multiplier to everything, which means if you bring the contrast down it squishes the already pretty subtle differences of the mid levels down to even less distinguishable. My guess is Hauppauge is applying contrast logarithmically, this helps it maintain detail while shrinking the overall range of the image.

    TL;DR because the Hauppauge is on-board and can take direct advantage of my processor I think it's able to do more advanced math which produces better results.

    Now, I am even more curious to look at how a proper analog proc amp would affect the image these two capture cards receive, but I don't have that hardware currently. My belief is the GV-USB2 would benefit more than the Hauppauge, but it's almost certain both would benefit.

    I feel like I understand what's going on well enough to at least finish that video I'm working on. Thanks again to everyone who piped in, will continue to be open to thoughts/discussion on any topics included here.
    Image Attached Files
    Last edited by ArcanaXIX; 15th Jun 2022 at 23:34.
    Quote Quote  
  10. Member
    Join Date
    May 2005
    Location
    Australia-PAL Land
    Search Comp PM
    Thanks for the effort, ArcanaXiX.
    Quote Quote  
  11. Originally Posted by Alwyn View Post
    Thanks Arcana, I will view with interest. I normally set the levels with the histogram before capturing and then don't adjust them again.
    Sorry about the delay, took forever to get around to recording the narration and editing/processing. It's kind of long, and again this particular video is focused on game capture rather than video digitizing, but if you're still interested:

    https://www.youtube.com/watch?v=dDChnrwHh-Y
    Quote Quote  
  12. Member
    Join Date
    May 2005
    Location
    Australia-PAL Land
    Search Comp PM
    Thanks ArcanaXIX, that's a very good and thorough explanation. I did notice you using the mouse to adjust the Procamp. I normally use the keyboard arrows to fine-tune, which I find gives me more control over the levels on the histogram.
    Quote Quote  
  13. oh wow, TIL. it may not be shown in the video but I was able to do as minute changes as possible (point by point) by clicking adjacent to the bar, but arrow keys are going to make my life a whole lot easier. Literally didn't even try that. Thanks for the tip!
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!