Capturing Video8 tapes: resolution, filtering, codecs, audio?
Hi everybody! Since I have my whole setup all ready to go, I hope it's alright I made a new thread for my questions involving capturing.
I'm about to start capturing my old video8 tapes in virtualdub avi huffyuv. The reason I'm capturing like this is because I want to use these files as masters and don't want to have to go back and re-buy my equipment after I sell it off if I want to edit the clips. I have done a few test captures on 5 minute tape I threw together and am pleased with the results. My last capture I had 0 dropped and 0 inserted frames, audio came out nicely and in sync. This is where I'm stuck: I hear this response both ways and am unsure which route I should take. I don't know if I should capture in 352x480 or 720x480. My test capture has been in 720x480 and it looks good to me, but from someone more experienced, should I choose one over the other? I may edit down the road which is why I was leaning towards 720 but not sure how necessary it actually is since the native res is 352. Space of my captures isn't a problem. Filters: I am a bit hesitant about using filters from virtualdub as I don't think I'm experienced enough to know what to use and not use. I'd actually prefer how the tapes are, but if I opt out of using virtualdub's filters is there a way or another program I can use to get equal results via post processing instead of during the capture? Do I need to tweak any settings in the configuration of the Huffyuv codec? I did follow the guide as closey as I could but want to make sure its setup how it supposed to be. Under YUY2 compression method I have it set to: Predict median (best). RGB compression method: Predict gradient (best). Field Threshold 288 lines. And the only box I have checked is: Enable full size output-buffer (might fix crashing). I have an AVT8710 I bought off ebay used, I know there are settings on the box I can change but I don't want to. Do these boxes retain any previous settings or do they revert back to default when powered off? I can't really tell from the footage if there's any color adjustments from the TBC. Next question is about audio. I have a santa cruz internally connected to my aiw 7500. Is there a specific "sweet spot" on the audio mixer that I should be adjusting to? I have it around 75% and when I preview the levels they get close-ish to the end but don't seem to clip. Will this setting vary per tape? Maybe I should post a test capture, is there anything specific I should record on one of the few blank video8 tapes that would help you guys figure out if I'm doing anything wrong? Thank you guys so much for all of the help, I never thought I'd get this far :) |
I hope there aren't too many typos below -- getting late here.
First advice: Make one capture as a practise cap. Don't try to get them all at once. It's more tiring than you think, and no one likes having to do them over again later. Quote:
Capture to 720x480. Quote:
Quote:
Quote:
Quote:
Quote:
Audio: There's no universal "sweet spot" for audio volume. Yes, they will usually vary by tape, even if a little. 75% output volume is about where most people start, just watch for high-frequency peaks or harshness. Because most PC speakers have harsh or "hashy" high ends anyway (or they have depressed HF response that's very deceptive), you should make adjustments using decent headphones. Don't use earphones. I use Grado SR80e and SR60e headphones for low-output devices like PC sound cards. SONY phones are harsh, Bose phones have no dynamics and an artificially hollow sound (unless you want to spend several hundred bucks for SONY or Bose Pro studio phones, and they're still overpriced). Quote:
|
1 Attachment(s)
sanlyn, I appreciate you reading everything I wrote and your reply so much, thank you! :)
Sorry for such a late reply, got busy with school. I don't know why but I kept thinking that the only color correcting that could be done was during capture. Now that I know that's all done in post mostly, I'll most likely toy with some filters and see what I can do with my footage. I've attached a clip, hoping I did it right and it's enough to tell if my settings/equipment are doing the job, and also am hoping the footage looks up to par so I can continue capturing the rest of my tapes! (btw, the footage does have audio but I muted it as it may have screaming and I didn't want to hurt anyone's ears :laugh:) Thanks so much! Johnny |
Sorry for the delay, and thanks for the sample. I'm traveling to visit elderly relatives this week and have only a small Netbook with me, which is not suitable for working with video. Others might comment before I return home this week, but for now I see that brights in the image are clipped and exceed the usual bright-end limits for digital video. Some of that detail can be recovered later, but you'll need very basic Avisynth to do it without losing highlights. I'll offer some pointers on doing that with VDub capture when I can prepare something on a real PC. The capture looks otherwise OK.
|
No problem! Thanks for the observation on my clip. Are the brights being clipped due to how the video was originally shot or was it due to my capture settings?
|
6 Attachment(s)
Quote:
For a first sample capture, you're ahead of many beginners. If you're using denoisers for captures, I'd suggest not doing so. If you didn't use noise reduction when making the sample capture, we'll need another explanation for the problems mentioned below. The 7500 didn't seem to cause any problems, nor did it ever caused me any headaches over the years. I guess I'll have to play devil's advocate here, thanking you for a good first effort on one hand, and bringing up some very common and not-so-common problems on the other. The video has a ragged, noisy look with buzzing, fuzzy, sawtooth edges, and shifting vertical waves of grayish banding and repetitious ghosting during motion. It will probably look cleaner on a CRT, but buzzing edges, blocky shadows, and ghosting will still be prominent. The frames referenced below are deinterlaced frame numbers. For illustration, the frames were deinterlaced temporarily using yadif in Avisynth. Later, for final MPEG output, QTGMC was used instead of yadif. deinterlaced frame 10 = interlaced original 5 deinterlaced frame 62 = interlaced original 31 deinterlaced frame 146 = interlaced original 73 deinterlaced frame 188 = interlaced original 94 deinterlaced frame 252 = interlaced original 126 Below, full scene image from deinterlaced frame 146 (side borders and head-switchng noise at bottom removed to avoid affecting the histogram). The YUV histogram at the right shows luma (top white bar in the histogram) overflowing into the unsafe video area beyond RGB 255 at the right side. In the histogram, darker side colors indicate unsafe areas: video-safe areas between the unsafe borders indicate RGB 16-235 for the luma video standard. Unsafe clipped brights are seen in most frames thru the course of the video. http://www.digitalfaq.com/forum/atta...1&d=1446526393 Below, a 2X blowup of detail from deinterlaced frame 146, which is interlaced frame 73 in the original sample. The noise is mostly coarsely mottled and smeared grain, and noisy bleeding red is so saturated that red detail and contours are obscured. The red "splatter" in the lower right is chroma ghosting. Red also tends toward purple here, and the video has an overall slight cyan cast. http://www.digitalfaq.com/forum/atta...1&d=1446526532 Below, a 2X blowup taken from the upper left-center of deinterlaced frame 146. The Red ghosting is chroma smear from earlier frames as the camera pans up and to the right. Ragged white sputter is seen on many contrasting edges, appears to be remnants of DCT ringing and edge halos. http://www.digitalfaq.com/forum/atta...1&d=1446526646 Below, smeared chroma noise and ghosting in deinterlaced frame 252. http://www.digitalfaq.com/forum/atta...1&d=1446526695 Below, multiple edge ghosting in 3 deinterlaced frames: frame 10 (left), frame 62 (middle), and frame 252 (right). Most of the detail in the lamp at the left is clipped beyond RGB 255 -- some clipping came from too much contrast in the capture, the rest was caused from trying to photograph a wider contrast range than video media can handle. The edge ghosts are probably caused by temporal denoisers used during playback or capture. Some of the bright level detail can be retrieved with processing in YUV. The ghosts can't be repaired. http://www.digitalfaq.com/forum/atta...1&d=1446526766 The original video shows poor interlace behavior from either the original camera or the player, shown as strong combing artifacts and aliasing. The noise in the video is coarse low-frequency type -- removing it completely would result in more banding and block noise during encode. The smeared chroma and color noise, the type of coarse shadow grain, and combing and aliasing, are defects that require deinterlacing to clean, then re-interlacing to maintain smooth motion. The swift motion requires a high bitrate and small GOP's. The yadif deinterlacer was used in Avisynth to create the posted images. For the attached mpg Preview_480i.mpg, the deinterlacer was QTGMC with light denoising enabled. Bright levels and some of the color bleed were tamed with ColorYUV, SmoothLevels, and awarpsharp2. The camera's autowhite and AGC caused mild but visible and measurable fluctuations in gamma and color balance. I didn't address those fluctuations, but they can be partially smoothed with Avisynth plugins. Other denoisers were TemporalSoften, GradFun2DBmod (anti-banding), and Santiag (anti-aliasing) in Avisynth, with CamcorderColorDenoise in VirtualDub. The video was reinterlaced in Avisynth for DVD and BluRay. 8 pixels of lower-border noise were cropped off and the original frame size was maintained by adding 4 pixels of black border at the top and 4 black pixels at the bottom, thus centering the image vertically. Most of this processing can't be accomplished in VirtualDub alone. The MPEG encoding used TMPGEnc Plus 2.5. If one encodes captures with defects like these without some cleanup, the result will be noisy, annoying video with all the glitches that are present in the capture, which means that most of the encoding bitrate will be used to encode glitches. In VirtualDub capture, black and bright levels can be managed with the "Levels" filter group, which hooks into the 7500's proc amp controls. Bright(ness) raises or lowers blacks, contrast raises or lowers brights. The two controls interact somewhat, so some fiddling is needed but it's easy to get used to. You should use VDub's capture histogram feature to check your results: when the histogram goes "red" at either the left or right side, levels are out of range. Remember that black borders will always be a red peak or "spike" at the far left-hand side. Regardless of the software you use to process video, it's neither practical nor possible to work without some kind of histogram -- which is simply a tool to keep your eyes from lying to you and for graphically telling you what's going on. |
(EDIT) Double post
|
I had to rename the thread. Undescriptiveness is a pet peeve. :)
720x480 is best for lossless AVI. Only use 352x480 for direct-to-MPEG (DVD) capturing. Then again, 352x480 lossless datarate is much smaller. And Video8 is at or under 352x480. So if the card is capturing that resolution well, and the video doesn't really need restoration, it's an option. Saving space is sometimes needed. Just be aware that not all cards do well at anything other that 720x480 max capture resolution. (Edit: It seems you have an ATI AIW, so you should be fine. But again, test it.) I've done this for post-restoration clips in the past, for archival storage. You can't filter (in VirtualDub) when capturing anyway. So save filtering for your non-master copies. Leave Huffyuv at defaults. I never use the weak AVT-8710 controls eithers. (Hopefully the AVT-8710 has no defects -- many post-2010 models do.) The sweet spot is always 50% threshold volume. All across the board. Disable any in/out not in use (mic, aux, etc). 75% is much too high. Hardware filtering via proc amps is what tweaks colors, and there are special boxes to filter for noise (and even I don't own those). For hardware, the most important items are the S-VHS VCR with line TBC and the external frame sync TBC. Never deinterlace your masters, only the copies. VHS natively looks terrible. You can do better. Start with a good VCR and TBC. That's half the battle. After capture, you can tweak chroma defects (shift, shimmer) and noise. It's not hard to make a VHS tape look almost DVD-quality, with resolution being the only giveaway. For consumer analog, Hi8 was best, followed by Video8, then S/VHS/c comes in last. Hi8 tends to be fairly flawless, but Video8 has a lot of VHS-like issue in chroma (smearing, offset), as well as cheap cameras to make it. The ghosting errors I see hear look like typical Video8 camera work -- assuming this is a straight capture, not a re-edit. That ghosting error is actually one I can somewhat tolerate. |
The glitches I saw are fairly common except for the level of ghosting. I'd point to cameras for that and would think that it's on the original tape. Some cleanup of other problems does make it more viewable. Hopefully most of the other videos don't have this much excessive camera motion. If they do, it will require high bitrates all the way.
Overall, it still looks better than my first captures way-back-when. I'd post a sample, but it's too embarrassing. :o |
2 Attachment(s)
The question often asked: is this cleanup work necessary? Do I have to learn anything about video? Can't I just click some icons and encode the thing?
The answers to the above are no, no, and yes. It will likely look suitable on a small PC monitor. Show the same vido on your 37" or 60: HD TV, and it's a different story. Send off that noisy video to YouTube or FaceBook, and get the shock of your life after net streaming techs get their hands on it. Viewing those 'net videos on tiny internet players is one thing. But download it and play it full screen on your PC...... The advantage of lossless capture is that it's as close to a 1:1 digital representation of your analog tape as you'll get without going into the kind of gear and software used by Disney or Industrial Light & Magic. And even then, those guys do some NASA-level cleanup you won't believe. Below are some before and after comparisons of the original AVI sample and results of work done with it. Mind you, I used some common fixes that were fairly easy to set up. Things would have been far more complicated had I really been serious. But most hobbyists, even those who want to do a little filtering, seldom get so serious. The results of moderate cleanup are a cleaner, smoother running video that's easier to encode and easier to watch. Below are images of the left-hand half or so of two frames. Just to keep things more fair, the images are from interlaced frames, not from progressive frames. Your media player, set top player, or TV will deinterlace these frames during play. Because they're still interlaced you'll see the double-image effects -- along with any improvements from moderate filtering. Interlaced frame 5: original AVI (left side), after cleanup (right side): http://www.digitalfaq.com/forum/atta...1&d=1446653517 On the left side you'll see excessive combing artifacts and "buzz" on edges. Bright parts of the lampshade are clipped, showing unnatural hot spots and some loss of detail. Even on a PC monitor, that hot spot will be clipped. A TV won't be able to accept it: the hot spot will be clipped and will likely change color, becoming more blue or green or some other off-color. The shadow area on the wall is coarse-graned stuff that eats up data bits on encoding. In a video with less motion, such shadows will look clumpy, grimey, and even more noisy. The right side shows smoothing of combing noise and rough grain; you see the double-edge effect of motion and interlacing, but edges are more well defined and less disturbed during play. Even the ghosting has been smoothed out somewhat. The shadows look more realistic. Lowering brights in Avisynth and YUV tames the lampshade hot spots a bit and retrieves a little more detail, though some of it is still lost during the original in-camera recording. Interlaced frame 31: original AVI (left side), after cleanup (right side): http://www.digitalfaq.com/forum/atta...1&d=1446654055 Check the ghosting and uneven effects in the walls on the left side, compare them to the smoother versions on the right. Edges of the foreground furniture are also cleaner. The blue-white-red object on the right has cleaner edges and less cholor noise, and the two bright panels in the dark door in the background are more distinct. The edges of the walls and picture frame are also less blurry. Taming the bright end stopped those white stripes in the foreground furniture from giving that "glowing" neon effect that marks so many videos with poor level control. One could probably go bananas and use some fancy and time-consuming techniques, but I decided that some moderate work had reached the point of diminishing returns. Obviously the ghosting won't go away entirely, not even with obsessive motion-interpolation that would take forever to calculate and set up. It would hardly be worth the effort. The filters I used are all mainstays, nothing exotic, used everywhere by everybody. The workflow is simply to open a video with Avisynth, load whatever filters are appropriate, monitor the results in VirtualDub and even add some VirtualDub filters if you like (I added Camcorder Color Denoise), then save the results as a new AVI and send it to the encoder. There's nothing unusual about this kind of restoration. If you browse thru the restoration area of this forum you'll see video so bad you wouldn't believe it was possible for that kind of stuff to exist (but it does, believe me). :) |
Wow! Had a super busy week but so happy to see all of this great information sanlyn and lordsmurf! I really cannot explain how appreciative I am for the time put into your replies, this is unlike any other forum I've participated in. So thank you :)!
@sanlyn, post #6 Quote:
I did not realize all of the issues with my tape, I think I was too caught up in seeing footage from a couple of decades to notice, now that you've pointed it out I can see the problems. Quote:
Quote:
@lordsmurf, post #8 Quote:
Quote:
@sanlyn Quote:
So after all the info you've provided me with, I will definitely be doing some restoration since it doesn't seem near as difficult as I had thought and the results you've provided look fantastic! Now, I'm left with this thought, should I continue to capture my tapes how the one I posted was captured? Or is there anything I need to double check that could be causing the issues appearing? I just want to make sure what's happening is because of how the tape was recorded, and not something I could prevent. Again, thank you guys so much |
3 Attachment(s)
Likely you didn't apply denoising during capture, but in any case it's not a good idea. Some highfalutin' VCR's have denoising built in. In many situations users turn it off or tone it down. Older DNR methods are primitive compared with post-process filters today. Besides, many cap software denoisers make colorspace conversions you don't want, and they slow everything down.
Quote:
I haven't had problems with audio clipping. I do get some that's kind of schreechy (3500KHz peaks), but filters in the free Audacity app can usually tame it. My audio input stays at about 75%. You can always increase volume later if needed. Some people make restoration processing look too easy or too difficult. Obviously it's a big mystery the first time you look into it, but after a few forum examples you start catching on. Every tape is different, but all of them have problems in common -- those that I pointed out aren't unusual, so a handy collecton of specific filters for cleanup will get used again and again. Quote:
one. Good riddance, I say, those I have now are better anyway. Aliasing and noisy interlace are common glitches with consumer cameras unless you happened to get into 4 figures when buying one. Everyone does the best with what they have. Quote:
Below are some examples, not one of which is a perfect capture ready to encode. The first is an underexposure from a home video, partly caused by bright lawn furniture giving fits to the camera's AGC, and partly the result of my sister habitually letting her camera's battery run down to nothing. Later in this same segment, which has a lot of pans and zooms, brights get blown off the map. So I set levels to just "contain" the worst damage, and fixed details later. http://www.digitalfaq.com/forum/atta...1&d=1446930230 The images are interlaced. Pretty dark scene. During capture I set the worst blacks and brights to stay inside RGB 16-235. As you can see in the left image, RGB display makes darks look darker. During this shot, levels changed about 15 times. The corrections (right-hand image) started with a luma offset in YUV to open up darks. The rest was fixed in VirtualDub. I didn't use "brightnes" or "contrast" filters, which are limited and which you'd be stuck with when using the typical consumer NLE. This was done with ColorMill and gradation curves. Both of them are free add-ons for Virtualdub and mimic similar controls in the high-priced "Pro" apps. If you look close, you'll see some of the same aliasing on motion that you see in most consumer cameras. What a headache. Below, two examples from a current project -- transferring an old VHS of Fantasia to DVD. Here's proof that even retail tapes are headaches, especially if it's a bad and aging film transfer from the 80's. The frame below has a common tape problem -- a rip or partial dropout right across the middle of the frame. Actually, it lasted for two frames. This sort of junk is common with tapes. In the right-hand correction you can see I still have a problem with some edge halos, meaning I still have work to do. No sharpening was used during capture or processing. Avisynth has a few special filters for that problem. http://www.digitalfaq.com/forum/atta...1&d=1446930297 I had to use an Avisynth motion-interpolation plugin to get rid of that glitch. Pretty simple to set up, really. It has only 1 setting. Below, a typical problem with the murky way analog tape renders shadows. The original capture levels were set to avoid some dense below-zero blacks and burned-up brights later in this segment. The tape was played with an AG-1980 and captured with an ATI AIW 9600XT AGP. Even with decent gear you still need tweaks. http://www.digitalfaq.com/forum/atta...1&d=1446930355 The corrections were made with a simple contrast-mask filter in Avisynth and YUY2 (again, only one setting which I had set to very low power) to bring up just a little more detail out of the darks. The rest was done in Avisynth with the same color filters mentioned above. After you see enough problems like this, corrections get rather routine. By the way, the small white dots in the darks aren't spots. It's pixie dust that floats everywhere. Everything I did used many of the the same filters I used on your sample AVI. Quote:
|
Site design, images and content © 2002-2024 The Digital FAQ, www.digitalFAQ.com
Forum Software by vBulletin · Copyright © 2024 Jelsoft Enterprises Ltd.