View Full Version : VHS OVERSAMPLING. An idea whose time has left.

Mutley Eugenius
09-09-2009, 11:40 PM
I have used a trick a couple of times when converting someone's VHS tapes to DVD, I was lucky enough to have more than one copy of the recording. I captured them both, synced them, made one 50% transparent and it makes for a much cleaner final clip, using the best single soundtrack. It was pretty cool.

Once I even did a 3 copy capture and it was an awesome final.

You have to allow for positional differences and use some XY correction on one of them. There also may be some framedropping issues over longer clips.

If you duplicate one of the clips and push it one frame out of sync you can get even cleaner images, but you get a mushy motion. But sometimes it's cool.

09-10-2009, 12:23 AM
It would probably look, even better if you did your process on clips enlarged to a HDV timeline and then reduced the size after rendering it.

Shawn Farrell
05-16-2010, 08:59 PM

Mutley Eugenius
06-02-2010, 03:40 PM
Actually Shawn, it's halving or third-ing the subtle white noise effect and signal dropouts that come from the tape it was distrubuted on. Analog tape is an imperfect medium (not that any medium is perfect) because it doesn't make a perfect duplicate, but if you have two or 3 tapes overlayed, having different imperfections, blending them will reduce the imperfections by 50% or 66% and you get a much nicer final product.

I know what a phase control does.

06-03-2010, 12:12 AM
That's an interesting idea. I might try it with some of my older stuff.

06-05-2010, 12:12 PM
If you just took the blue channel of one of the captured tapes and overlaid it, you would just get a lot of the detail that is blown out form lighting and are really only in the red and green channel. Of course not so easy to utilize this method in SE but something I have done in Fusion (you can do it in AE as well). But you could try removing the red and green in the control tree and then do the 50% transparency. I think I tested this once and it worked not as well as in a compositing program.

06-09-2010, 02:55 PM
...of course [this "blue color channel blending trick"] is not so easy to utilize this method in SpeedEDIT...

...[yet it's] something I have done in Fusion (you can do it in After Effects as well)....

... But [in SpeedEDIT], you could try removing the red and green [via] the control tree and then do the 50% transparency. I think I tested this once and it worked not as well as in a compositing program.

You provoke within me, Josť, a nagging question:

How is it that SpeedEDIT is *so* unlike many compositing programs -- other than the absense of all those cool transparency modes?

(Is SpeedEDIT's "lack of transparency modes" issue, the major factor as to why the above "Video Enhancement by Additive Blue Color Channel" trick doesn't work so effectively with our beloved NewTek editor and its predecessor, VT-EDIT?)

I've been fascinated by this enigma for some time now. That said, thanks Josť -- and any other expert(s) who might provide additional insights on SpeedEDIT vs. Compositing programs.

06-09-2010, 03:42 PM
I like this idea. I think it has merit.

06-09-2010, 08:21 PM
SE2 does not have transfer modes so all we have is the keying and alpha to work with. A compositing program and some other editing programs have transfer modes like screen.
Anyway since SE2 does not, the best way to enhance footage is to look at the waveform - RGB scope in the master color tab and you can see where detail is on a per channel basis. Bright or blown out spots will usually be so high in the red and green that they are clipped but the blue will not because of the way the color spectrum is truly seeing the bright spots. Obviously this will not work on totally blown out 8 bit video but I have found more detail in the blue channel even in those extreme cases.
Sometimes just lowering the red and green gain in the waveform -RGB or in the control tree is enough to lower the hot spot and show the detail in the blue channel. Also just placing a copy of the clip and have the copy just have blue gain and then overlay and lower the alpha.
Anyway just load a clip that has back-lighting like a sunset and look at the blue channel in the waveform - RGB to see clearly what I am referring to.

Take care,

06-10-2010, 01:49 AM
Fascinating stuff. Thanks Josť. I used to relish Sony Vegas 5 for its transparency functions.

I'm curious: Would you say AE is easy to grasp as a comp tool than, say, Fusion? TIA

06-10-2010, 08:38 AM
First let me say that I don't want my NKE being a compositing program. I would like some transfer modes though and the ability to use an alpha from another layer to matte a layer.
Now as far as AE and Fusion, they are not really each others competition IMHO. Fusion is a high end compositor costing $5390.00. AE included in the CS5 Production suite is $1699.00. I own Fusion 5 (not upgraded yet since my client base has not warranted it but I would love the tools in version 6) and I own AE CS4. Before AE CS4, I did all my compositing in Fusion. Then when I purchased CS4 Production suite, I started to play with AE again (I used AE about 8 years ago and then switched to Fusion) and for motion graphics and some other things, due to it's vast included plug-ins, I have found it to allow me to punch certain work faster.
But back to the question, Fusion is nodal and AE is layered. I take nodal over layered any day of the week. It was a little weird when I first learned it but once I got it, I loved it. As a sidebar, learning nodals helped me quickly jump into LW nodals when they were introduced. So I really don't think it's fair to compare the two applications.
I'll also add there are thousands of AE tutorials and reference material floating around and you would be able to learn AE and continue learning it with out issues.

If you edit in SE2 and need a section of your project to use compositing tools in AE, just open a 2nd SE2 and copy and paste the section or clips you want to send to AE into the 2nd SE2. In the 2nd SE2, save project as a .avi. Now open AE and import the .avi wrapper file. This involves zero rendering since the .avi project is pointing to the project files via a wrapper. Now do you AE work and then sadly, render out of AE (but at least you did not have to render out of SE2). Now you can load the new rendered clip out of AE into the space the original came from and you have a work flow to allow the use of AE while working with SE2.

Take care,

06-10-2010, 08:48 AM
I'd like transfer modes in SE as well, but just as a reminder, those with VT's can apply very similar 'merge modes' in Aura (or its siblings) when they need them and don't want to shell out for a compositing app.

06-10-2010, 09:18 AM
I'd like transfer modes in SE as well, but just as a reminder, those with VT's can apply very similar 'merge modes' in Aura (or its siblings) when they need them and don't want to shell out for a compositing app.

Yes very true Steve. I made the mistake of limiting my response to SE and for that matter SE2. Assuming a person only had SE 1 or SE2 and not VT-SE.

And correct me if I'm wrong Steve but when you use a transfer mode in Aura, isn't it destructive?

06-10-2010, 09:24 AM
And correct me if I'm wrong Steve but when you use a transfer mode in Aura, isn't it destructive?Yes and no. It doesn't really have to be destructive - for example, you can merge without removing the original merged layer, keeping it as a backup, hiding it, etc.

But this isn't the same case as working in a real compositor (or even Photoshop), where you can set a layer to a transfer mode without merging, and see the result, tweak it, and so on, up until the time you flatten it. In Aura, the result of merging is basically like flattening the merged layers, apart from deliberately keeping 'un-merged' copies of the layers for re-use later.

06-10-2010, 09:28 AM
OK right it ask you the option to keep the original layer. But like you said, if you have it and you are on a budget, it does do transfer modes :)

Take care,

Mutley Eugenius
07-15-2010, 03:11 PM
The VHS Oversampling idea is also a trick that can be used very successfully with chromakeying. I have done this one a few times with grainy bluescreen or greenscreen footage and it makes the difference between rescuing the chromakeying of the footage or trashing it and starting again...

1) take the footage into SE and separate it from it's audio.

2) slow it down to 50%

3) make a duplicate of the slow clip and put it right below on the time line (holding down the v key for vertical alignment)

4) select the lower clip & press Ctrl+> to move the lower clip one frame to the right.

5) press f8 & set alpha to 128 for the lower clip, and press y on the clip to make it overlay enabled.

6) if you play the clip and look at it you will notice that you have now blurred one field into the next. This gives a slightly mushy motion to the video, but is serves a great deal to average out the differences between pixel brightnesses and thus it dramatically reduces grainyness and frame to frame jitter of chromakey edges. If you are willing to make a sacrifice of a bit of quality, you can also apply a minor blur to one or both of the layers (experiment with small numbers on this one, a little goes a long way but too much will hurt your clip), which also serves to remove grain by averaging neighboring pixels.

7) render the clip, at 200%, which will restore it to it's original speed, then check it out.

A note on focus blur, vertical blur is more effective for most codecs because codecs often use rectangular pixels for color information. Vertical blurring will smudge the chromas of 3 lines into one line. If you used horizontal blurring you would have to blur about 5 pixels wide to get the same or similar effect with rectangular chroma pixel codecs. For more info on this, check into the definitions of '4:2:2' and other similar ratios.

You can also get into the nuances of blurring individual channels if you want to and seeing how green channel blurring or blue channel blurring affects your chromakey's algorithm.

If you want to go really wild, blur the hell out of your clip and paste it over the original, make it very transparent to only influence the colors a bit, and render it again. Then see what you get.

07-15-2010, 07:25 PM
Nice trick but SE2 has new slow motion modes that could make this a little eaiser Mutley.

Mutley Eugenius
07-15-2010, 10:23 PM
You're probably right, I imagine that slowing it down to 50% and then setting the slow mode to linear probably does the same thing if that's how the algorithm works... only a programmer would know for sure.