[Lars Bunch]

Hi,

I mostly work in HD but frequently provide files in SD. I prefer to work with the largest frame I can (In most cases, 1920×1080, square pixels). If the footage was interlaced, I usually try to remove it if I can before doing any real work. Generally the frame rate of the original footage will dictate the frame rate I work at, although given a preference, I like 24fps or 23.98. If I am working at 23.98 or 24, I stay in that frame rate until I’m ready to output an SD version with interlacing and a 3:2 pulldown.

Generally I try to stay in the HD world and only render to SD as the final step. As my last step, I will usually render an HD file in the Animation codec and then re-import it, modify the output levels to match 601 colorspace and add a 3:2 pulldown to get it up to 29.97 if necessary.

HD gives you more colors to work with than NTSC as well as more pixels. Tracking is easier in HD since you have more detail to work with, although as with SD, you will need to remove interlacing for tracking.

But your render times will be a good deal longer. If your comps start out as SD you can work much faster, but only on fairly simple projects can you easily modify the comp up to HD resolution. If there is a good chance you will need to provide an HD version, then you are best off starting in HD since the time involved in modifying an SD comp up to HD is often longer than the render times.

There are plenty of different comp presets you can choose. I like to stick the the simplest one (for my purposes) which is the 1920×1080 SqPix, 23.98 HD. By working with square pixels, I don’t have to deal with the oddities of different pixel aspect ratios until the final SD render. Since the footage I usually work with is 1920×1080, I work at that size, but if you are getting a smaller HD format (1280×720) you probably won’t see any benefit to working larger. As I mentioned, you’ll probably be best off working at the frame rate of the footage you receive. I prefer 24 partially because it is easy to add a 3:2 pulldown to get it to 29.97fps and also because I don’t have to work with as many frames. Believe me, if you do any rotoscoping, you’ll be happy to not have to deal with 6 extra frames per second.

If you have to provide a 4×3 SD version, you will need to protect the 4×3 area on any 16×9 animations. This usually ends up meaning your compositions don’t look good in either 4×3 or 16×9. Ideally you can work purely in 16×9 and not have to either compromise framing or create two variations of every animation.

For final output, you should render the HD animations with the Animation codec (or Uncompressed 422), re-import them and drop them into a new comp with the output size and pixel aspect ratio and frame rate that you need. In outputting for the web, you may want to just use quicktime pro (or windows media encoder) to export the HD file to whatever format you need.

There is a lot to consider, but I suppose it can be boiled down to keeping everything as large and unmodified from the original format for as long as possible and only shrinking down and changing frame rates as the last step. The only exception is if you have interlaced footage, try to get it deinterlaced as soon as possible.

Okay, this is a bit long winded, but it’s really not that difficult. When I started doing HD, I too was concerned about all the details, but you’ll find it’s pretty straight forward once you get used to the differences between the formats.

Hope this helps,

Lars

[Lars Bunch]

Hi,

What is the difference in quality that you are seeing? Are the colors different? Are they muted? Are the levels different?

AE 5.5 does not have any color management features (at least not like CS3) and so if the Photoshop file has an embedded color profile that is being interpreted for display in Photoshop, when it is opened in AE, that color profile will be ignored and the colors may look completely different.

You might want to start by opening the file in Photoshop and converting it’s color profile to sRGB IEC61996 or Adobe RGB 1998.

If your photoshop file is in CMYK, you will see really extreme color problems in AE. You should convert any CMYK files to RGB before working with them in AE 5.5.

If this doesn’t solve your problem, please give us more information about what you are seeing and perhaps someone will have the right answer for you.

Hope this helps,

Lars

[Lars Bunch]

Hi,

I use both Mocha and SynthEyes, although my experience with either program is only in the past few months. The major difference is whether you need to track in 2D space or 3D space. Mocha exports tracking data in 2D space and is output is intended to go into a compositing program where 2D planes are layered over one another. SynthEyes exports 3D data including a camera move and is designed to work with 3D animation programs.

I would choose Mocha if I needed to corner pin a 2D plane into a shot or if I needed to roto a person. If the effect I need to do needs to exactly match movement already in the original shot, the side of a moving bus for example, then Mocha can create the data that I need to build the composite.

Syntheyes is more useful if you have a moving camera shot and you need to place a 3D object (or a 2D plane floating in 3D space) into the 3D space of the original shot. This is useful if you are adding elements that need to move independent of anything else within the shot, but they need to be seen from the point of view of the camera that shot the original footage.

Yesterday I needed to replace the text on a shot zooming in on a welcome mat. Mocha tracked to mat perfectly and I used the corner pin data to drop in the new text. This involved a plane of replacement text overlayed onto the original shot.

A few days ago I needed to fly some text through a dolly shot in a hotel lobby. The text sweeps past our POV and around a pillar, arcs around behind a couple of people and swings around to rest above their heads. Here I used SynthEyes to obtain the 3D camera move that I could import into Maya. At this point it became a simple matter of animating the text along a curve while letting the camera move wherever SynthEyes told it to. Because Syntheyes reconstructed the original camera move, all I had to worry about was creating the movement of the text itself. Once I rendered the text, I could easily composite it with the original shot and it appeared to move within the 3D depth of the scene.

You can download and experiment with Syntheyes’ 3D tracking since it allows you to drop in 3D object to see how well they track. The tutorials on the site are very helpful.

When you export a camera from SynthEyes to AE, you will see the most benefit in 3D aware plug-ins such as Invigorator or Trapcode plugins. If you do full 3D animation, you can do much more complex interactions with the original shot.

If you do a lot of 2D tracking for AE, I strongly recommend Mocha. I have never been happy with AE’s trackers and Mocha really fulfills that need.

I have not experimented with it, but apparently SynthEyes can export the 2D tracking data for AE so, to some extent, it can do a lot of what Mocha can do, but Mocha cannot give you a 3D camera move so if you could pick only one, SynthEyes may be the way to go.

But if you do a lot of 2D tracking, Mocha can do it better since that is what it is optimized for.

Ideally you have gobs of cash laying about and you can get both programs.

Hope this helps,

Lars

[Lars Bunch]

Hi,

In the comp window, there is a little button at the bottom that looks like a rectangle with arrows above it. It is next to the number of views pulldown in CS3. This toggles between a square pixel representation of the comp and an interpreted pixel (the way it ought to look.)

Lars

[Lars Bunch]

Hi,

Is there any problem in rendering to DVCPro HD? When I have supplied animations that have needed to cut into a project done in DVCPro, I have rendered in that codec and have gotten good looking results. I have found it to be a pretty good codec for high compression with little artifacting.

I usually try to render to the codec that the editor is working in. If the shot needs to cut into an Uncompressed 4:2:2 timeline, that’s what I’ll render to. If it is DV, I’ll usually render to Animation codec since DV looks bad no matter what you do. But DVCPro generally looks pretty good.

Lars

[Lars Bunch]

Hi Again,

There are a lot of ways you can change the color. An easy one is the Hue/Saturation effect. Drop that into the dress layer and click on the “colorize” check box. Then you can adjust the hue, animating it to to transform to a different color.

You might also want to try the Tint effect. If I remember right, this allows you to change highlight and shadows separately.

If you want to get really weird, you could experiment with the Colorama effect.

Hope that helps,

Lars

[Lars Bunch]

Hi,

Quicktime 7.4.1 seems to fix the problem, but because of the dire nature of the fiasco it would probably be a good idea to do a full back up of your boot drive before upgrading anyway.

At any rate, I ran into the bug while running 7.4.0 some time after 7.4.1 was available. I immediately downloaded 7.4.1 and was able to continue working.

This was on a Macbook (intel) running OSx 10.4. I am now running 10.5.2 with Quicktime 7.4.1 and all seems to be well.

Anyway, all seems well, but proceed with caution.

Lars

[Lars Bunch]

Hi,

After you’ve keyed the footage (using keylight, for example) just render your footage to a codec that supports an alpha channel (such as “Animation”)

In your render settings, you will have to make sure it is set to Millions of Colors+ The “+” indicates it will add the alpha channel.

If your client is editing on an Avid, you should render using the “straight” option rather than “pre-multiplied” Actually straight will work for most uses, but the quicktime file will look really ugly if it is played on the desktop. It will look okay in the Avid. Aharon Rabinowitz has a good tutorial on this issue.

https://library.creativecow.net/articles/rabinowitz_aharon/straight_vs_premult.php

If you need to provide the footage on tape, you will need to render out the main footage and then an alpha only version that you can then record to tape. The editor will have to marry the footage with the alpha down the line, so you might want to add a sync mark at the head of the footage and the alpha. (This is a pretty unlikely scenario so I wouldn’t plan on it… I just add it to help confuse the issue.)

Hope this helps,

Lars

[Lars Bunch]

HI,

I don’t have an answer, but I’ve seen the same problem. In fact, I recently saw the problem for the first time on an animation that seemed okay a couple of days earlier. I am wondering if this might be related to a version of CS3 (I’m running the most recent version [8.0.2?]) or open GL. Actually I didn’t see the problem when working on a G5 or Macbook, but when I upgraded to a MacPro, I noticed it. It might not be related, but that’s when I saw it. This makes me wonder if it might be related to the graphics card.

I didn’t see the problem in the final render so, at least in my case, I assumed it was a preview issue. I assume you are seeing it in the final render on your animations.

What version of AE and Quicktime are you running? (Major and minor verion numbers.) What is your OS and your graphics card?

Where I have seen this sort of problem more is in Maya when the resolution of a light’s shadow settings is too low. There doesn’t seem to be a comparable setting in AE, but do you see any changes to the quality of the pixelation if you adjust the light’s shadow darkness and fall off etc?

Lars

[Lars Bunch]

HI,

I don’t know if this is particularly easy, but it should be pretty straight forward…

Shoot the dancers against the green screen with their dresses in whatever color you need. Key the footage as you normally would. Render the keyed footage in full color with an alpha. Create an inverted version of the alpha. Render an Alpha only quicktime and re-import it as footage.

The rendered alpha becomes your silhouetted dancers. Put that into your comp. Drop the footage of keyed dancers on top of the silhouette. Rotoscope the dresses, dropping out the stuff you don’t want. (Depending upon the amount of movement and length of time, this could get really old very quickly.)

You might be able to cheat by using a tint effect to make the shot a single saturated color so you can get a bit sloppy on the roto without it showing. If each dress is a different single color you could make as many layers of the color shot as you have dancers, tinting each as needed and then rotoscoping each of them separately. Come to think of it, it might be better to do that anyway. That way if you run into an error with one mask, you are less in risk of messing up previous roto work on the same layer. It’s also a good idea to lock finished layers so you don’t accidentally modify them later on.

If you tint the shot, it is best to do that after you’ve done your rotoscoping. Actually if you temporarily boost the color saturation of the shot you are rotoscoping, it will help show you the edges and may make the work a little easier. You can then delete the saturation effect and replace it with the tint effect when you are done.

If the dresses have very clean colors you might be able to extract them using a key, but you would have to do a separate key for each color of dress. I’m not sure how exactly this is done in AE since my experience at keying is mostly in Shake, but it can be done.

If you are going to do a lot of roto, if you don’t have a Wacom tablet consider getting one. Actually the Cintiq 12″ is ideal, but a bit pricey.

Good luck,

Lars