Bill Lee
Forum Replies Created
-
Yes. Sort of. Maybe. And sometimes Not.
The Difference Key is the filter you would have to use here. This filter allows you to mask out any pixels in a frame that are the same as in a reference image (your “Clean Plate”). You find frame that has none of the moving people in it, save it as a still picture, re-import it as your clean plate. You add the Difference Key to your video, open the filters tab and drop your clean plate into the image well. Adjust the threshold and tolerance and it all works, except….
Where it doesn’t work:
-
If you start off with high resolution still images for an animation, then it may be easy enough to use Photoshop to create masks for each frame. The high resolution will help because the later resampling (maybe) to SD will mitigate the effects of differently selected pixels if you were working with SD material with Photoshop. For HD output, the contribution of individual pixels is less obvious for the overall look and so it can be acceptable for doing it this way.
My big concern for you is that what seems easy for a couple of frames can become a nighmare for hundreds of frames if doing each frame manually. The use of shapes and keyframes in rotoscoping is meant to avoid manually working on each frame wherever possible and let the rotoscoping software provide the tweens for the rotoscope shapes between keyframes. Working on curve points every couple of frames instead of every frame dramatically reduces your workload and make the animation smoother.
I’d have to disagree with an acceptance of poor rotoscoping in stop motion photography. There is no need for poor rotoscoping even for this art form, despite the limited number of frames you typically have to work with. In other words, the distinctive animation look of your stop motion should come from the stop motion and not from inadequate rotoscoping. You may want vary the rotoscoping to achieve a certain look which is fine, but is is great to have that option and not be forced to have that look.
Bill Lee
-
I’d have to agree with Arniepix about Shake being a good option. Unless your rotoscoping is very simple, the frame by frame rotoscoping in Photoshop can lead to a flickering effect on the fringes of your rotoscope. This is because you are treating each frame separately, rather that being a series of related frames. A slight difference in the masks you are creating for each frame will cause this effect, where rotoscope shapes created and changing over a number of frames should have better smoothness and fluidity.
Shake is very affordable these days (compared to what it used to be) at US$499 and less if you are a student/academic staff. The question is: is it worth the expenditure of money and is that going to pay off in terms of quality or savings in time. Only you will know that for the project you are working on. The more rotoscope you are doing, the greater the benefits.
When creating rotoscope shapes from curves, a good suggestion is to build your full rotoscope out of the union of a number of simpler shapes, due to the non-independent nature of the curve to each point on that curve. Thus instead of having to tweak, say, a lot of points in a curve outlining a whole body, you just change the points that make up a leg or part of a leg which is easier and gives better results.
Picking the smallest number of good keyframes and curve-defining points is an art in itself: too many points and it’s a nightmare to get the rotoscope smooth, too many keyframes and you risk a flickering effect like you might get from rotoscoping in Photoshop.
Bill Lee
-
I’ve been doing some more thinking on this, and in the Photoshop help for CS2 there is the following note:
By default, type is displayed with the fractional widths option on. This option allows the spacing between characters to vary using fractions of whole pixels. In most situations, fractional character widths provide the best spacing for type appearance and readability. However, for type in small sizes (smaller than 20 points) displayed online, fractional widths could cause type to run together or have too much extra space, making it difficult to read.
Turn off fractional widths when you want to fix type spacing in whole-pixel increments and prevent small type from running together. The fractional character width setting applies to all characters on a type layer
-
OK, I set up a multi-layer Photoshop file (.psd) 720 x 486 (D1) and imported this into a D1 8-Bit uncompressed sequence to do some tests. I think I may have discovered why you are getting the results you have been getting.
A multi-layer photoshop file is storing the layers as a set of the smallest bounding boxes for each layer. This is true for both raster layers and for text layers.
If you have a 10 x 10 pixel square on your 720 x 486 D1 frame, then it is stored as a 10×10 box with an offset from the origin of the D1 frame. You can see this when you import this multi-layer psd file into FCP, drop it in the timeline, then double click on it to see the layers. If you turn on ‘Image and Wireframe’ in the Canvas view, clicking on each layer will show the bounding box of each layer. Since each bounding box is not centred on the frame, each bounding box will likely have a different offset from the centre of the frame and thus show a different set of values for the centre point in the Motion tab. The same is true for text layers, the text being stored with the bounding box for the text and the centre point of that bounding box.
Text layers are somewhat special, since you have to be careful about font metrics and rasterisation when moving to other applications or machines. When you move a text layer to FCP, there may be some subtle differences in the way the text is rasterised for compositing into the frame. Photoshop gives you the ability to anti-alias text and these have an effect on the way the text is displayed. These anti-aliasing settings will have an effect on the size of the bounding box when it is imported into FCP, and this is why your text may be changing position.
The test that I did:
-
Apologies to all.
Yes you are right, you can’t add FCP-recognised Chapter markers during a capture with Live Capture Plus 2 (LCP2).
The statement from their website: “mark clips and events while a capture is in progress.” only puts a timecode into the notes field when you press Command-T “Insert Timecode Marker” and then allows you to type anything after that. You can put multiple timecode markers in this notes field. If you export the captured file as XML, it doesn’t make these ‘marker’s into real markers, but puts them in your clips notes field as one concatenated string. Obviously it would have been nice to have these exported as markers and then imported into FCP as markers, but Live Capture Plus can’t do that right now. I’ve just suggested it to the developers as feedback.
zrb123’s posting prompted me to download the demo software and have a closer look at it and markers in FCP.
The XML file exported from FCP for a clip with markers generates a definition inside the <clip> and </clip> pairs, for example – a marker at ten frames in from the beginning of the clip with a duration of 4 frames:
-
When you import still pictures into FCP, it looks at the sizes of the images you are importing and makes some decisions based on those sizes. On images that it thinks are standard video images, it will set the pixel aspect ratio to one of: Square, NTSC – CCIR 601, PAL – CCIR 601, HDV (960×720), HD (1280×1080). If you are just one pixel out, it will default to square pixels.
Example: Importing an image 720 x 487 will import that image with a square pixel aspect ratio, whereas an image 720 x 486 will be imported as NTSC – CCIR 601.
When you drag such an image into your timeline, the image will be scaled if it is larger than the frame size. An example of this is that an image 721 x 487 was scaled to 99.86% to make the largest dimension (x or y) fit into the frame. Images smaller than frame size will not be scaled to fit.
If your sequence is 720 x 480 DV, then your pictures should also be 720 x 480 instead of 720 x 486 D1. Otherwise, set the scale to 100% and let the three extra pixels hang off the top and bottom of your frame. Likewise, if your sequence is D1, your pictures should be D1 (720 x 486), otherwise you will get a three pixel black line top and bottom of the frame (immaterial for TV, bad for web delivery).
So as to your situation, did you scale the picture in Photoshop, or in FCP? I can’t remember running up against it automatically offsetting an image before, since the default position of images and video is at (0,0) which is the centre of the frame.
If you change the pixel aspect ratio of some clip in the timeline, then you have to replace it with another copy of that (modified aspect ratio) clip from the browser, since it won’t visibly change aspect ratio once it is in the timeline.
Bill Lee
-
Jeremy “The display allows you to toggle, but it warns you it’s for display purposes only.”
The message that Photoshop CS 2 gives you is “Pixel aspect ratio correction is for preview purposes only. Turn it off for maximum image quality.”
I think this is there just to warn you that you are not dealing with square pixels, and that there may be round off errors with some wierd aspect ratios in displaying an image or selecting pixels. The advantages IMHO greatly outweigh any possible disadvantages with not having to rescale images multiple times if you are round-tripping between Photoshop and FCP. The only area which I don’t know whether it handles the non-square pixels correctly is in filters that use radii of effect, such as the Gaussian Blur: when applied to an anamorphic image, it should give an effect which is circular instead of oval-shaped. I don’t know if it actually does or not.For so many years of Photoshop, all pixels were square and now you can set them to an anamorphic ratio. When you have the pixel aspect ratio turned on, you can now get guaranteed circles using the Eliptical marquee tool in conjunction with the Shift key without later having to rescale. A pencil tools one pixel square looks rectangular onscreen, but is actually one pixel wide and high. In essence, that’s why Adobe made these presets available: to be used to edit frames for video.
If you look in the help for Photoshop CS 2, it has the following information:
“To help you create images for video, Photoshop has a Pixel Aspect Ratio Correction viewing mode that displays your image at a specific aspect ratio. For more accurate previews, Photoshop also has a Video Preview command that lets you immediately preview your document on a display device, such as a video monitor. To use this feature, you must have the device connected to your computer via FireWire. See also ‘To preview your document on a video monitor’.”Bill Lee
-
This was the old method of using Photoshop.
As of Photoshop CS onwards, you can select non-square pixels as the Preset for a new document. This means that you can work with 720 x 480 NTSC (DV) or 720 x 486 NTSC (D1) frame sizes with Pixel Aspect ratios of other than 1:1. They even have frame sizes for anamorphic 16:9 video.
This means that the process of round-tripping between FCP and Photoshop is vastly improved, since you leave the frame size constant and work with non-square pixels when drawing. You can toggle the display between non-square and square pixels in Photoshop with View>Pixel Aspect Ratio Correction.
Bill Lee
-
Boot from and run the hardware test disk that came with the machine. It may be poorly seated or marginal RAM that only gets used when you are using FCP. If the problem only occurs after a while, then it could be related to the heat inside the computer causing RAM (or logic board) failures. Either way, run the hardware test overnight running looping RAM tests to see if, when booted off the CD/DVD, the machine fails. If it does so, it eliminates the possibility of your software having been corrupted.
If you’ve got a spare hard disk, install a copy of the latest copy of the OS on it, together with the minimal amount of FCP/S to do testing with your material. This should take about an hour or two at most for the install. Run with this minimal software and see if you still have problems. Running the latest version of FCP would be best, since you never know what bugs may have been present and squashed by a later update of software. This way you won’t have to interrupt your existing setup. I’ve done this on one of my iMacs since I was getting kernel panics and poor performance (it gets used as a TV set with an EyeTV tuner attached) – and now I’ve narrowed the problems down to the software on the iMac since everything has been fine running the temporary software off an external FireWire drive for the last three months.
Bill Lee