ReelSmartTM TwixtorTMUsers' Manual.

NOTE: We have tutorial projects for commonTwixtor uses. See our website, http://www.revisionfx.com. Go to the "Tech Support" section andnavigate to the support page. In particular we support

People that are already familiar with Twixtor1.4 or before can refer to our techsupport section for Twixtor to clarifywhich parts of this document are changed or are new in 2.0.

What is ReelSmartTMTwixtor?

ReelSmartTM Twixtor intelligently slows down, speeds up orchanges the frame rate of your image sequences.

ReelSmart Twixtor enables you to "time warp"your image sequences with visually stunning results. In order toachieve its unparalleled image quality, Twixtorsynthesizes unique new frames by warping and interpolating framesof the original sequence... employing RE:Vision's proprietarytracking technology that calculates motion for each individualpixel.

ReelSmart Twixtor works in Adobe(R) After Effects (R), Apple (R) Final Cut Pro(TM),  Adobe Premiere 6.0 (R) or later, PinnacleCommotion 2.6 (R) or later,  Discreet (TM) Combustion (TM) ,and eyeon Digital Fusion (R) (via their After Effectsadapter) and other programs that support After Effects plugins.

Twixtor Features:

Twixtor Pro adds the following features:

Important Note: Because of application limitations,specifying different frame rates for your input and outputsequences is only supported from within After Effects.

Before You Start, EXTREMELY IMPORTANT.
Common problems discussed here!

For all applications

PLEASE READ: The document "How to set up host applications for RE:Vision Effects plugins that work with fields" describes what you need to do in the host application in order to use ReelSmart Twixtor.

After Effects

Twixtor Pro and After Effects 5.0 Pro version.

Combustion

Commotion

Final Cut Pro

Premiere 6.0 or later

For any Premiere projects using Twixtor:

If you'd like to make a clip shorter (speed it up) by a factor of 3 with Twixtor, then you'll want to make a Virtual Clip that points to the entirety of the original clip. Then edit the endpoint of the Virtual Clip so that the Virtual Clip's duration is the shortened duration length (for example, if the clip is 3:00 seconds long. Make a Virtual Clip that points to the original clip. The Virtual Clip will then be 3:00 in duration. You then set the duration of the Virtual Clip to 1:00. Then apply Twixtor to the ORIGINAL clip, and set the Constant Stretch Factor to 0.33.

Okay, in one case Twixtor is applied to the Virtual Clip, and in the other, to the original clip. How do you decide which clip to apply Twixtor to? RULE OF THUMB: Always apply Twixtor to the clip that is of longer duration.

Note: Twixtor actually changes the frames that are displayed. However, when you walk along the timeline in Premiere, frames are generally displayed pre-Twixtor. To see the Twixtored version, you'll need to Render Scrub (see the Premiere manual), or better yet, Preview the working area to RAM or disk (again, see your Premiere manual).
 

Digital Fusion

You have to make sure that both inputs are connected. Twixtor actually uses the one that corresponds to the TimeWarp Layer. And naturally you need the After Effects plugin adapter.

Have you already read the manual and want a parameterreference (complete with links to their descriptions)? Here yougo!

Usage of ReelSmartTMTwixtor

Color Source: Color Source specifies the clip to be "Twixtorized."

Color Source must have some value other than "None!". The Color Source clip is tracked and it's color information is used to create new frames. Note that another clip can be specified with Alt Motion Source for tracking purposes, limiting the role of Color Source to be the RGB values that are warped. Until you get familiar with Twixtor, simply leave Alt Motion Source set to "None."

Frame Interpolation: A menu option with 3 choices.

  1. Nearest: warps the nearest calculated input frame by the right amount to make the current output frame.
  2. Blend: Performs bi-directional warping and then blends the resulting two warped images.. The more you stretch a clip, the more this choice is recommended.
  3. Motion Weighted Blend: When a new frame is created from existing frames it is often the case that some softening is introduced. This softening can more or less strong (depending on how far away in time the new frame is located from existing frames in the original sequence. This can result in a pulsing of what appears to be the Twixtored sequence coming in and out of focus. The Motion Weighted Blend option attempts to reduce this pulsing.

Motion Vector Quality: A menu option with 5 choices. With "High Quality" set:

  1. No Motion Vectors: no motion vectors are calculated
  2. Sloppy: motion vectors are calculated at 1/4 resolution, then interpolated. Furthermore, a faster but less accurate method is used.
  3. Medium: motion vectors are calculated at 1/4 resolution, then interpolated.
  4. High: motion vectors are calculated at 1/2 resolution, then interpolated.
  5. Best: motion vectors are calculated at full resolution, then interpolated.
  6. For "Draft " mode we calculate motion vectors at half the resolution as specified above.

For demonstration purposes, let's assume that we are scaling a clip's duration by a factor of 3. As such, output frame 25 would conceptually need input frame 8.33333. How is frame 8.333 calculated?

Note that if fewer frames are being generated for the output sequence, that is, the input sequence is being speed up (total time duration is shrunk) OR a lower output frame rate than input frame rate is specified, then motion blur may then be applied (see Motion Blur option below).
 

Crossing objects? (Conflicting motion?)

Sometimes the motion estimator will perform better on footage that has multiple motions in the same layer if this box is checked. "Multiple motions" occur when 2 or more objects pass by each other in different directions, or at differing movement rates. Note that this option increases motion vector calculation time significantly.
 

Motion Blur: 

When speeding up a clip, you are essentially telling Twixtor that there is "more action" happening per frame. In order to simulate what happens in a real camera, and to reduce flickering, Twixtor applies motion blur when speeding up footage in order to simulate what happens when more motion takes place in the same amount of (virtual) shutter opening of a camera. You can dial in how much motion blur is added with the Motion Blur slider.

Note that the application of motion blur in Twixtor only occurs when footage is sped up, to compensate for different virtual shutter speeds. This is different from our ReelSmart Motion Blur product, which adds motion blur based on the motion calculated in a sequence without speeding up or slowing down the footage.

Specifying fields and input/output FPS rates.

What are fields?
1) Simply put, an interlaced video frame consists of 2 fields. One field consists of all even lines of a frame, and the other field consists of odd lines. A video device (monitor, TV, etc) has a beam that traces, or draws, one field, then goes back and draws the other. As such, one field is displayed first, then the other is displayed half of a frame time later.

2) So interlaced video has two fields per frame. One field is displayed first in time, and then the other. If the top scan line is part of the field that is displayed first (that is, the field of odd scan lines is displayed first) then we say that the video is "Upper Field First." If the second-to-top scan line is part of the field that is displayed first (that is, the field of even scan lines is displayed first) then we say that the is "Lower Field First ".

The rule of thumb when specifying "field first" for your footage: if you choose lower or upper field first and the plugin produces material that is overly jittery, or jumps back and forth, then you probably chose the incorrect option.

Note that you cannot assume all of your source material will have the same field first-ness.

Changing sequence timing.

Okay, now here's the meat of the plugin. How do I retime my footage?

First let's assume that the input frame rate is the same as the output frame rate.

Stretching Method  pop-up menu has 3 options for the retiming method, as follows:

If Constant Stretch is chosen, then Twixtor assumes that you'll be changing the duration of the clip by the amount specified in the Stretch/Speed parameter (formerly called Constant Stretch Factor). For example, if you want a 1 second sequence to be 2 seconds in length, then set this value to 2. (duh!). Note that if you have a 1 second clip that you want to make 2 seconds then you'll have to first put your 1 second clip in a composition that is 2 seconds long (without any filters). Then you'll want to apply Twixtor to this 2 second long composition inside another composition (see the "Before You Start" section above). This is because filters are not allowed to extend the length of the layer it's being applied to. If Constant Stretch is chosen, the parameter should NOT BE ANIMATED.

To use the Frame parameter for retiming:

Immediately after applying Twixtor to a clip, it is advisable that you:

First, make the Frame parameter animatable.

Second, you are advised to make keyframes for the Frame parameter as follows:

  • make a keyframe with value 0 at time 0
  • make a keyframe at the last time value in the sequence, with a value equal to the last frame number
    •

    For example: Let's say you have a sequence of 300 frames.

  • make a keyframe with a value of 0 at time 0
  • make a keyframe with value 300 at time 300 (time, in this case, is in "frames" instead of timecode).
    •

    This will give you an identity mapping (frame 0 returned at time 0, frame 300 returned at time 300, frame 47 will be returned at time 47, etc.).

    The third menu item lets you animate the speed of the input clip. To obtain the speed of the clip, Twixtor gets the value from the Stretch/Speed slider. A value of 0.5 tells Twixtor to play the clip 1/2 as fast as the original, a value of 3 tells Twixtor to play the original clip at 3 times its normal rate. Unlike the Constant Stretch method, the Speed parameter MAY BE ANIMATED!

    Stretch and Speed are very much opposite specifications. Setting a Constant Stretch value of 3 is the same as specifying a Speed value of 1/3.0. The main difference is the following: if you try to animate the Constant Stretch slider you will get unpredictable results. The Speed value, when animated, gives very intuitive results.

    If Stretch Method is set to Constant Stretch or Speed then the Frame parameter is ignored. Conversely, the Stretch/Speed slider is ignored when key-framing destination frames. TIP: If you are doing a complex dynamic remapping of the video action, you can quickly set your key-frames by setting Motion Vector Quality to None and Frame Interpolation set to Blend. When done setting keyframes, you can go back and set the Motion Vector Quality to the desired quality.

    EXERCISE: If you are reading this manual for the first time, you should try taking a movie you have lying around and try to expand or contract it's duration using Constant Stretch, keeping the output frame rate the same as the input frame rate.

    Note for After Effects Only:  Different Frame Rates

    Okay, but what if the input and output frame rates are different? If you don't want to change the duration of the clip, (and you are using After Effects) you're in luck. Simply set the input and output field rates as appropriate and your done! The default settings of Twixtor will do the right thing for you.

    If you want to change the duration of the input source at a constant rate, make sure Stretch Method is set to Constant Stretch. Twixtor automatically accounts for the differences in frame rates. The Constant Stretch Factor scales the total number of seconds. If you want to convert 1 second of 24fps footage into 2 seconds of 30 fps footage, set Constant Stretch Factor to 2. Again, the Constant Stretch Factor parameter scales the total number of seconds of a clip. In order to get a handle on this, think of Constant Stretch Factor as changing the total number of seconds and not the frame count. (In our case here 1 second of 24fps footage, or 24 frames, is converted to 2 seconds of 30 fps footage, or 60 frames).

    Note for After Effects Only: Remapping Time with Different Frame Rates

    What if you want to dynamically remap time AND change frame rates at the same time?
     

    It's a little tricky to keyframe time on top of a frame rate conversion but it can be done. The keyframing should be done presuming that the input has already been converted to the output frame frame. To make this concrete: say you have a value of 60 for Frame, and the output frame rate is 30fps and the input frame rate is 24fps. What input frame is retrieved? We first calculate the time value of the frame in the output frame rate (frame 60 = 60/30fps the frame at 2.0 seconds). Then we convert this 2.0 seconds time value to the input frame by multiplying by the input frame rate (2.0*24fps = input frame 48). Of course fractional frames are retrieved, warped and blended as described before.

    Notes for Interlaced Video Usage Only: Handling Fields

    Okay, you ask, what happens when there are fields in the input (interlaced video)?

    There are two ways to approach this problem. Method 1 is how we suggest you first try Twixtor.

    1) Let Twixtor create frames from fields using field interpolation. This is accomplished by setting "Fields" in the Twixtor settings to "Upper" or "Lower" as appropriate and making sure that the application you are using passes Twixtor images with both fields (i.e., interlaced footage). When passing Twixtor footage with fields, each field in an input source becomes a frame to be interpolated. So 30i footage really looks like 60p to Twixtor. A standard algorithm is applied to make each frame (we interpolate the fields to create a whole frame). Input frame 0.0 is made from the first field (in time, which may be upper or lower depending on your source footage) of the first frame. Frame 0.5 is made from the second field in time of the first frame and so on. Make sure to set Input:Fields to Lower or Upper, as appropriate, in the Twixtor settings.

    To set up Twixtor so that Twixtor receives frames with both fields (some applications try to be "smart" and sends deinterlaced frames to Twixtor) please read: "How to set up host applications for RE:Vision Effects plugins that work with fields."

    2) You may decide that Twixtor's field interpolation for making frames for a particular piece of footage does not do the job for you, and you would prefer to use some other deinterlacing process and pass Twixtor full frames. Make sure to set Input:Fields to None in the Twixtor settings.

    NOTE: To see Twixtor's internal deinterlacing, set Twixtor's Stretching Method to Constant Stretch and set Stretch/Speed to 2.0. Set Input:Fields as appropriate and Output:Fields to None. This will create a frame from each field of your interlaced video.

    NOTE 2: Twixtor will perform better if better deinterlacing is performed before the application of Twixtor. Our ReelSmart FieldsKit plugin set contains such a better deinterlacing plugin, which tries to create better edges, and only performs field interpolation where motion is detected. The difference will be most significant in shots where there are portions of the image that is static from frame to frame (like when the camera does not move).

    For interlaced output, frames at 1/2 frame times are calculated (and warped and blended as you specify) and then interlaced.

    I'm slowing footage down, do I really need to recalulate the motion vectors every frame?

    The answer is no!

    Plenty Memory? (Cache Last Motion?):

    If you have plenty of memory, check the Plenty Memory? (Cache Last Motion?) checkbox. This will cache the last motion vector calculation. This will only be helpful when slowing down footage (and is ignored when speeding up footage). If your footage is 720x486, then you'll need 2 extra image buffers worth of memory (or 8*width*height bytes), or if you are using Twixtor Pro with a foreground mask, you will need 8 extra images worth of memory (=16*width*height bytes) if you are using Twixtor Pro with a foreground layer mask.

    Warning: If after you've checked Plenty Memory? you change the clip used as the Color Source or Alt Motion Source or FG Mask, please make sure to turn off, then turn on this option in order that this option will function properly. (by turning the option off and on we make sure to clear to motion buffers based on the new clip. None of the applications we plug into notify Twixtor when you change your clip selections.

    The tracking is not working so well on my source footage, what can I do?

    Tracking problems show up as motion not being properly tracked: either objects in the output sequence "gloop" together, or objects don't move enough causing what seems to be simple cross dissolves of the input images. (for what constitues problems for our tracker, see the "PROBLEMS IN INPUT SEQUENCES" section below.

    Alt Motion Source:
    You might say, "if only I could enhance the colors used in the tracking process, but then use the colors from the original sequence for the Twixtored sequence."  Or, the reverse, you've heavily filtered the original (which you've used as the color source for Twixtor), but you think you might get better tracking results by using the original footage as the imagery to be tracked. So, we've provided this parameter to set an Alternate Motion Source: If Alt Motion Source is set to None then the Color Source is used for both tracking and the color of the Twixtor sequence. However, if an Alt Motion Source clip is specified, then this sequence is used for tracking and the Color Source clip is used for color. Of course, Twixtor assumes that both the Alt Motion Source and Color Source clips line up exactly in time.

    Motion Sensitivity:
    Okay, you might say "I don't have time to experiment with image processing my footage to specify an alternate motion source," or worse, "I try to color correct my footage and it does not help the tracking!" If objects still "gloop" together when they shouldn't, then you can try manipulating the Motion Sensitivity parameter. Motion Sensitivity limits how much pixels can move. A value of zero assures that pixels can't move very much, and a value of 100 allows pixels to move as much as the motion interpolator can calculate.  The default value might sometimes be a bit too "ambitious" for some material... by reducing the sensitivity you might be able to create a more interesting result (giving a more traditional frame blended look) where there is large motion rather then some sort of unappropriate swirl that Twixtor might introduce.

    So, how can you enhance tracking using an Alt Motion Source? The following set of pictures seeks to explain:

    Figure 1

    Figure 2

    Figure 3

    Figure 4

    Using an alternate motion source:

    I have a matte for a foreground object and would like to improve Twixtor's tracking by using this matte (or, Improved tracking Part Two, Twixtor Pro Only)

    Note that if you do not have Twixtor Pro, you can still use this feature in demo mode (the word "Demo" will appear before the FG Mask option, described below).

    Twixtor Pro allows you to specify a mask so that you can track a foreground object separate from the background.This can help immensely with "gloopiness" in the Twixtored sequence. Except for specifying nice anti-aliased edges (and perhaps incorporating some motion blur) of the foreground, make sure that mask is full-on white. Be careful for example of not having tiny holes in the supplied mask.

    Please note that while using a foreground mask can dramatically improve results, Twixtor needs to calculate one or two additional motion estimations per frame (as we now generate motion for the foreground, and motion for the background), so it can take more than twice as long as without a foreground layer.

    In order to describe the Inv Mask Shrink option, pictures are worth a thousand words! So let's say we have the following image, and matte for the foreground object:
     

    Source Image

    Matte for Foreground Object

    Using the matte for the foreground and the inverse of the matte for the background we get the following two pictures:

    Matte used to separate out foreground.

    Inverse of matte to specify background.

    Note that while the matte specifies the foreground quite nicely, some of the pixels of the foreground spill into the background! By specifying how many pixels to shrink the inverse of the foreground matte (the FG Inv Matte Shrink value), we can reduce or eliminate the spillage. By specifying a value of 1.5 we get the following picture:
     

    Shrinking the inverse of the matte to get a better background by specifying a FG Inv Matte Shrink value of 1.5.

    Note you might ask, "won't that create a gap in my final output?" The answer is no! Through clever filling and compositing, we track the background separately from the foreground, and fill in missing information in the background as best we can!

    Using a foreground layer and memory usage.. Twixtor Pro will use approximately 6 image buffers' worth of memory to run the foreground layer option. (image buffer's worth of memory is approximately 4*width*height bytes, or if in 16-bit, then 8*width*height bytes), in addition to whether or not Plenty Memory? is checked.

    PROBLEMS IN INPUT SEQUENCES

    If you have not used Twixtor yet, this section mightprove to be a little bit esoteric. This information will be ofmore use to you as you become more familiar with Twixtorand want to understand how to generate the best possible results.

    It might happen that Twixtor might do poorly inspecific instances. The idea here is not to discourage you andtell you it does not work but, rather, to give truth inadvertising. As you become an expert user, you learn to predictthe kind of material that can cause problems.

    1.     "Picket Fence": A veryregular pattern in motion (for example someone wearing a t-shirtwith fine stripes) with an object moving in front of it (e.g. thesame person hand for instance) might confuse the motion vectorcalculation. Any very structured pattern rotating can cause"blind spots" for the analyzer.

    2.     "Transparency": Overlayof semi-opaque surfaces might create unexpected results. Somecameras for example will streak under fast motion and that cancreate disappointing tracking results.

    3.     "Short Interval Defects": Sudden global illumination change (eg flash), strobing, dust,... can create unexpected / undesired results. Also,  ifthere is a piece of hair or a scratch on the scanned film for aframe this would influence the tracking so you really should tryto clean such defects before processing.

    4.     "Fields": We provide somebasic field handling. Always remember that by definition, evenand odd scan lines in material with fields are considered half aframe apart timewise. Because fields are spatially a line offsetfrom each other, regions where there is no motion and horizontallines might  produce a slight up and down motion  inphase with the source frame rate.

    5.     "Duplicated Frames":Twixtor does not provide automatic duplicated frame detectionsupport. You have to be aware that if your material has 3:2pulldown you should remove it beforehand. The same applies toanimations on 2's.  This can look particular weird in aspeed-up when suddenly a frame would have no motion blur as aresult.

    6.     "Alternate Motions": Whenmotions going in different direction are layered it is possiblethat the dominant motion affects (spills into) the backgroundmotion.

    7.     "Specular Highlights": Ifyou have moving lights, e.g., a shiny object that reflects thelight as it moves, it might cause problems because when themotion estimator will attempt to match two images, it will tendto follow the highlight as if it was an object. Sometimes that iswhat you want though.

    8.     "Ultra-Fast StructuredMotion": We are very particularly perceptive to humanactions. We have sometimes seen that certain complex rapid motionsuch as someone doing a frenetic dance creates interframedisplacements that are just too big for our motion estimator toresolve satisfactorily.  When planning a shoot for an effectthat involves Twixtor in the pipeline, consider that for Twixtorfast articulated motion should be easier for front facing subjectthen sideway views as there will be less pixels traveled perframe on the screen, which is really the only thing that Twixtorcares about.

    9.     "Fast Cuts": Although itmakes sense why, we nonetheless repeat it, Twixtor might workbetter on clean sequence boundaries. That is,  if you applyTwixtor on some finished footage with fast-cut advertisingcontent or some MTV like action that has dissolves andmulti-layered of actions going in and out, it will do somethingbut the result might be more predictable if used on theunprocessed (before doing dissolves...) source elementsindividually (or you might prefer  to animate the MotionSensitivity in phase with your dissolve).

    10.  "Occlusions": Problems caused by objectmotions tend to be one frame problems and localized in an area ofthe frame which is called an occlusion, which is some pixels thatyou see on one frame but are not visible on the other as a resultof camera and/or that object motion.

    11. “Limited Reach”: As a rule of thumb considerthat Twixtor will be most accurate for pixel displacements thatare a maximum of 5% of your image resolution (for 720x486, thiscomes to a maximum horizontal displacement of 35 pixels or so).With displacements larger than that Twixtor will start to becomeless precise as it tries to separate motions from one another inan image sequence.

    12.  "Compression Artifacts": Certain videocoding techniques such as DV compressors use 8 by 8 pixels blocks(intraframe) based compression. What this means for you is thatif for instance you have a sharp edge that moves, it will switchof 8 by 8 block and therefore locally it's neighbor values willbe all different (substantially not like 2-3 values over 255 butsometimes 40 off near an edge).  This is why green screenlike setups perform badly with DV...  Without smoothing thesource, this can sometimes certainly create tracking problems.

    MORE AESTHETIC / COSMETIC CONSIDERATIONS:
    Gee! It breaks on these 3 frames what do I do?

    Now that we have discouraged you, let's help you to becreative in repairing problems. We know that most of the time,there is nothing we can do to prevent problems at the sourcebecause the material already exists, and you need to fix it. Solet us share a bit our experience. Experience we acquired playingwith this kind of process in production, in trying to make aproduct that works, and in supporting our users. There is anastronomic quantity of possible images and it's easy to developassumptions that prove not true if you always work on the samefew clips. So, so you know we fully appreciate short movies (5frames is usually enough) where Twixtor fails somehow, this ishow we learn, and can hope to make an always better product.

    1.     1 or 2 frame defects on sourcematerial:  Sometimes you receive a movie with one or twoframe defects (could be a flash for instance). Take the goodframes on both ends of the bad ones and slice them together intoanother comp, create a short Twixtor clip of the proper durationand reinsert it back over the bad frames. Naturally if you alsowant to Twixt the whole movie, perform the repair before applyingTwixtor to the whole sequence..

    2.     Splotches:  Although wetry to accommodate for separate layers with the Separate Motionsbutton there are times when you will still get some splotches(unwanted warping artifacts).  You can probably run Twixtortwice (once with Motion Vector Quality set to Best and once withMotion Vector Quality set to None, for example).  If you arescaling by a low-factor (under 1.5) then you might want to swapthe frames that are bad. If your problem is very local, you cancreate a matte to reveal a Tracking Quality None version fromwithin the Best version, using the compositing functions of yourapplication. You might want to do a soft matte for nice blending.And naturally if you have the PRO version, you can provide aForeground/Background separation mask to Twixtor which you shoulddo probably in a manner which respects the shape of the object inthe foreground.

    3.     Ghosts: Since errors such asghosting are usually due to occlusions (pixels that were notvisible in the previous frame, so don't have correspondence inthe other frame due to motion), they are usually easy to paintout so when the result is played in motion no one will see it.Some camera motion such as driving on a bridge tend to create amulti-frame ghost which usually is not too problematic to paintout in apps like Commotion or Combustion that have clone offsetpaint tools. You might also prefer to not apply motion blur atthis stage, paint/fix and then use ReelSmart Motion Blurto add some motion blur.

    4.     Using existing mattes:Sometimes, rotoing out a fast moving foreground object will makethe background very stable and produce tracking data without theforeground motion influence. If you have the luxury of havingmattes for a layer, and don’t have the PRO version, youmight want still to try to Twixt the foreground and backgroundseparately (so that the motion tracking influences on the layersgo away or are reduced by Twixting them separately). Note in somecases you would need to fill the foreground hole with some formof clean plate image so it has the general motion of it'ssurrounding area. This process is somewhat what the FG motionmasking services of the PRO version would do for youautomatically.

    5.     Shakes...: Although bydefinition we deal with camera shakes... note that when youslowmo something you will also slowmo the camera shakes andvice-versa, so you might consider to pre-stabilize the sequenceon which you want to apply Twixtor.

    6.     Multiple Segments: When youare applying Twixtor to a sequence that has multiple cuts in it,it might be preferable when doing large scale slow-motion toseparate the shots and apply Twixtor individually so as topreserve a sharp cut.  Also if you have dissolves betweentwo such segments, you will have to decide (if it's possible) ifyou want to do the slow-mo on the result or before, note you canalways turn down the Motion Sensitivity knob while you arepassing over  the dissolve segment or inversely you mightlike the fluid resulting transition.

    7.     Noise, compression artifacts:Our algorithms should not be too sensitive to small grain noiseat the analysis level, however it might look weird to extrapolatea lot of frames per frame on material that contains a lot ofnoise (as it essentially slows down the noise also) so you areadvised to remove some noise in the color source, perhaps runninga form of smoothing or noise-reduction filter before. Withmaterial such as some “real damage” MPEG 1 kind, loweraccuracy motion estimation might even produce more desirableresults.   Also with DV-like sources, avoid nearestinterpolation mode. The smoothing of blending actually reducesnoise.

    8.     Very Large SlowMo: Theamount of clean frames you can generate is often more a factor ofthe action in the scene as there is probably no answer to thequestion of how many frames can be inbetweened with a processlike this, as it depends... As a rule of thumb, after 5% of theimage size displacement for a single frame, consider yourselflucky if it works. (Again there is no absolute answers). Perhapsalso you should consider that there is a limit to how much youslow down things as even perfect inbetweening with a computeranimation systems can look weird as it won't have any localdynamics, that is it becomes similar to doing global shape keyframing, everything will sort of move at the same rate.

    9.     Softening: Another relatedthing to watch out is that you will notice that inbetweens will“soften” the texture details. The thinking is that it'ssort of OK as you don't need as much details when you move (youassume you will have some motion blur) and the image where itdoes not move will remain very clear. If unwanted softeningoccurs, and the Weighted Blend is not enough, you can considerlittle post tricks including slightly sharpening (maybe simply anunsharp filter) and adding a bit of noise... Depending on whatyou are doing this might be more or less realistic.

    10.  Handling Multiple Motions: A"rough" garbage matte roto that isolates the foregroundobject might be necessary on the few frames where the problemoccasionally occurs.  After applying Twixtor to the frameswith the garbage matte (on both foreground and background, withinverse mattes of course!), the inbetweened frames can berecomposited back together.  When you make a Motion Mask youcan typically be a bit more sloppy then usual roto forcompositing, a bit like color grading mattes as the result willbe composited back on itself, you are only helping Twixtor figureit out. In general since Layer selection is not animated it mightbe simpler sometimes to run the whole thing once and then animatesomething like the Motion Vector Quality so it’s really fastand then cut back the relevant parts of the two comps together.

    11.  Fast Action:  So, with fast actioncontent, you will probably get better results with Twixtor frommaterial captured with a 30i FPS camera than with a DV camerathat does 12 of 15 real FPS, because the same action will movefarther, per frame, in the 12fps footage than in the 30fpsfootage.  If you have the option of shooting progressive orinterlaced video, a basic rule of thumb might be if you have fastaction (or because you are running with a hand-held camera) shootinterlaced (because we construct 60fps from the 60 fields, whichgives us more frames to deal with, which reduces the motion perframe), otherwise shoot progressive (because we don't have toreconstruct a frame from a field, you will get better qualityresults).

    12.  Changing Field Order:  While Twixtorwill transform the footage from Upper to Lower field first (orvice-versa), you may want to consider doing the field orderconversion first, then applying Twixtor with the same input andoutput field order . The most effective way to change dominancewould be to drop the first field, and then reinterlace the fieldsstarting at field 2 of the footage (or alternatively shift thepicture by a line vertically). By doing the field orderconversion first you more likely reduce sampling artifacts byTwixtor's creation of whole frames from individual fields. Ofcourse, your mileage may vary :-).

    13.  Other Fields Related Problems:  IfFields cause you problems, remember that we offer also a full FieldsKitwhich allows to flexibly organize your workflow to possiblycreate a better deinterlaced source for Twixtor. Note that somepeople want to simulate a telecine look by going down to 24 fpsand then adding a 3:2 pulldown. This is certainly possible andFieldsKit will eliminate the need to create an intermediaryrender. But, it is our experience that you obtain somethingcloser to a film look by simply Deinterlacing and applyingReelSmart Motion Blur and in that case maybe you don’t needTwixtor.

    14.  Flickering:  Similarly weird flickeringas sometimes happen with problematic film capture can have animpact. Note Twixtor mostly uses the black and white informationand is by default hardwired to video brightness which largelydiscounts blue, sometimes just doing an invert  trackingsource will help. So particularly for shots which have a lot ofblue, it might be worth to supply an alternate track which remapscolors and pushes up certain values. This also applies withcertain dark shots (low-lights level), it might be worth toexpand their range and lift them up.

     

    ReelSmart and Twixtor are trademarks of RE:Vision Effects,Inc.
    (Adobe, After Effects, and Premiere are registered trademarks ofAdobe Systems, Inc., Apple is registered trademark and Final CutPro is a trademark of Apple Computer, Inc., Pinnacle is aregistered trademark and Commotion is a trademark of PinnacleSystems, Discreet and Combustion are trademarks of Autodesk, Inc.and eyeon Digital Fusion is a registered trademark of eyeonSoftware, Inc.)

    Copyright 2000-2001 RE:Vision Effects, Inc.