Visual Effects Techniques For Stop-Motion Feature




The fog is a character in the story. It is the Holy Company. It is of 3 types and we are not initially aware of its full power or sentient nature. At first it simply hovers along the ground creating atmosphere and obscuring distances. Later it acts with intent seeking out Ramon. Finally it becomes a collection of human forms – the spirits of all the people the Holy Company has captured.

1. Ambient background fog 2. Animated character fog 3. Spirit fog with human forms (Santa Company – Holy Company)

These fog styles should be designed as still images (Photoshop paintings) made using plates from the scenes where fog will be applied. These will be the art direction for all fog effects. The Holy Company figures should be modeled based on a practical figure. Photographic reference (lighting, motion, turnarounds) of this figure will be the best guide for achieving the desired geometry with CGI. The motion of the character fog may be difficult to pre- visualize but drawn animation will be useful for determining the flow and speed.

Much of the ambient and background fog can be created using paint layers with distortion effects (warp, twirl, pinch, etc.) applied to a multi-plane treatment in a compositing program. This technique should be sufficient to produce much of the bulk of the fog throughout the film.

The next layer of fog can be effectively animated and put into specific shapes and positions with a CGI simulation using a particle system. This will allow the creeping character fog to behave in very specific ways as it pursues Ramon. This CGI fog should be used only where necessary for this specific behavior because it should be quite expensive¬Ā for design, manipulation, and rendering, especially where changes are concerned. The 2D fog should be much more flexible and will take less time to adjust and alter.

The next layer will be the Holy Company fog. Animation is fairly minimal for these spirit characters. They float smoothly and never interact with the solid characters. What is important is their visual signature – their apparent density, they way they connect to the mass of fog, and the wisps of fog they leave as trails. Additionally they will need some internal turbulence like a self-wind reflecting their tortured and trapped state. Fernando has described this look as similar to the windy smearing in The Lord of the Rings when Frodo sees the wraiths while wearing the ring.

The best way to allow for the right look to be achieved will be to create a number of layers, and to develop the look in composite. While testing is the only way to determine exactly what these layers ultimately will be, a preliminary list might include the following:

1. Painted 2D fog with distortion
2. CGI fog.
3. Varying degrees of density for the spirits themselves from a thicker solid core through various stages of very light turbulence reaching the a light cloud based on the solid volume of the character.
4. The mass of the character itself rendered as fog.
5. A trail element generated by the geometry of the character.
6. A flowing foggy matte pass used to elegantly connect the spirits to the masses of fog.

Sufficient variation in these elements should allow a very nice look to be achieved in composite. The internal turbulence will probably work nicely as a composite effect applied to 1 or 2 of these layers – probably the solid mass itself (#4 above) – likely the least important element as far as the foggy mass is concerned.

The final layer for all the fog scenes should be practical. Isolated wisps of fog with very dynamic and natural character should be shot by the VFX crew. The set should be entirely black, and black props and fans should be used to create reactions and dynamic movements. It is important to avoid hard edges on the props and all shots should be wide enough that nothing ever touches the edge of frame. These elements are ESSENTIAL to creating an excellent fog effect. They will be used sparingly in the final composite but their value will be immense. These elements will really sell the effect by adding natural behavior to fog composites which will often look OK but not completely finished otherwise. The added benefit will be in saved schedule time and man-hours which would otherwise need to be used to get the generated fog to an acceptable level of quality.


There are many candles in O Apostolo. Because of their sheer numbers, economic processes will be essential in keeping the scope of this task under control. There will be 2 main challenges to achieving good results with the candles:

The first challenge will be to get the right behavior from the flames. Candle flames respond to small amounts of turbulence in their environment. Thus when someone walks by, or when a door opens or a gust of wind comes through a window, the flame usually responds with a moment of shaking. Practical elements do this perfectly but can be somewhat of an editorial challenge when there are many candles in a shot – they all must move in synchronization but with small variations. If they all move exactly together the effect will be awkward. If there is too much variety the effect will not be convincing. Good results can be accomplished by using a library of similar elements and offsetting duplicates of the same flame by 1 or 2 frames. For the

number of candles in O Apostolo, a collection of 5 or 6 of each desired type of flame movement should be sufficient. These libraries can be generated with live-action or CGI techniques. I prefer live-action practical elements. Their behavior will likely be more believable and satisfying, containing many unexpected variations which will be elusive and time consuming to create with a CGI simulation.

The second challenge with these effects will be placement in Z space. Generally the maximum horizontal alignment in a 3D frame pair (assuming 2K resolution) is somewhere between 20 and 30 pixels. Thus any discrete location in Z will be defined by a horizontal measurement which can be quite small and very exact. In a moderately deep 3D shot of 8 pixels for instance, a 1 pixel offset will move an object more than 10% forward or backwards. For this reason, sub-pixel positioning is often necessary to accurately marry a composited element to a plate at the EXACTLY CORRECT location in Z. This is further complicated by the fact that these minute gradations of Z location not apparent unless the composite is seen projected on a large screen in full color and at full resolution. This is likely to be a large an on-going challenge because there are many sequences with many candles each in O Apostolo.


There are a number of shots of people eating in the film. Practical or simulated live-action steam can be vey effective in a stop-motion context. This effect is quite easy and can be achieved in composite with mono live action elements. CGI techniques can be quite easy as well provided an acceptable treatment re: speed, turbulence, density, etc, can be found. CGI can be less desirable if the look is elusive or if rendering becomes cumbersome. Live action requires development of an element library – setting up an effects stage, and shooting, but the results are perfect for the desired application. A 3rd possibility is the use of stock effects elements. An internet search of stock resources may very likely yield an acceptable collection of elements.

These elements composite easily. Shot or rendered on black, they will produce the desired look with a mix (additive, screen, maximum) and transparency adjustment. Live action elements will need a contrast adjustment (levels – crush the blacks) to composite cleanly. A very soft mask in the inverse shape of the source of the steam (e.g. a mound of food on a plate) will connect the effect with its intended source.

ALL these composites should be carefully checked and adjusted often (on a large screen in full color and at full resolution) throughout the process to ensure accurate, pleasing, and professional results.


The challenge with the coins will be in the composites. 1. Their shiny surfaces should reflect the characters. This can be an easy composite task but correct stereoscopic placement will be very challenging. I suggest using the chroma- key elements as refection maps in CGI and generating a stereoscopically correct (relative to the coin surfaces) reflection pass. A cast shadow will probably be somewhat difficult also but the savings from having to do it in composite will be worth it. 2. To do the above, as well as to marry the elements may require the 3D to roto the plates. Shooting dimensional reference (cubes and grids) will help. CGI will also benefit from receiving all the camera data. This may be overkill because the layouts for this sequence suggest a fairly loose and forgiving treatment to the dimensionality, but the match of lens and roundness will need to be kept close enough for the composite to look right, and convincing shadows and reflections are essential.


There is 1 shot of a smoldering mound which will need a rising plume of smoke. This should be a fairly straightforward effect. Probably good to differentiate this from the fog by making sure the color is noticeably different.


There are some additional concerns with screen replacement type composites in stereoscopic films. Garbage mattes can have added ramifications in 3D because these will sometimes need to be independent in each eye. It is wise to keep a close watch on these elements as they are being shot, because fixing something on set can save many hours of double garbage matte work later.

Additionally, the chroma key elements will be mostly stereoscopic and occasionally will be shot with different lenses, framing, or interocular settings than the backgrounds they belong to. These variations between composited element and plate can lead to disparities in roundness and stereo depth. These can both be difficult to repair. Composited elements in stereoscopic 3D also present problems with shadows and reflections. Accurate placement of shadows and reflections can be difficult.

For these reasons, all chroma-key elements should be kept as simple as possible. FG and BG elements should be shot with the same setups whenever possible, and complex composites requiring shadows or reflections should be avoided whenever possible.


Rig and wire removal can be a simple and straightforward process in stereoscopic 3D. There will be some challenges. Rigs with Z-axis presence can require separate roto for each eye. Paint and shadow re-creation can be difficult to match in each eye – stereo paint disparities look bad and must be avoided. These often will only be apparent in a large screen projection. With stop-motion, clean plates will often not match making the composites which follow rig-roto more difficult. Here also stereo disparities will be very noticeable and need to be avoided by careful checking in projection.


Set shifts often occur in stop-motion. Smaller shifts can often be fixed with full frame transformations from scale & reposition, to corner-pin distortions and mesh warps. These fixes can usually be applied to each eye exactly the same with perfect results. Larger shifts can require Frankenstein¬Ā cut & paste techniques using pieces from an earlier or later portion of the shot, and clean plates. These types can often require individual attention to each eye of the stereo pair. Fixing set shift can be very tedious and frustrating, and can take many hours/days per shot. I would suggest tracking set shift carefully in an attempt to know at any given time, how much there is to fix, how much time the fixes are taking, and what percentage of shots contain shifts needing repair. This will give VFX a good idea (all throughout the schedule) of how many man-hours are going to be needed for set shift repair to complete the film.


Lighting pops often occur in stop-motion. They are usually no more of a problem in stereoscopic 3D than they are in 2D unless one is unlucky enough to have them occur in only one eye of the stereo pair. The technique for fixing these anomalies is the same in either case. Fine contrast controls of the expand/compress/gamma type (e.g. Photoshop levels) can bring the errant frames close enough to the rest of the sequence that the pop becomes invisible even if it is not entirely eliminated. Key-framing the correction and flipping between the previous and following frames will allow this to be fixed by eye quite quickly.

Lighting pops which affect each eye differently should be top priority and should be handled much more carefully. The reason is that this R-eye L-eye disparity (known as retinal rivalry) is very unpleasant to the viewer. The experience is subliminal and results in discomfort and general dissatisfaction with the experience i.e. people begin to dislike the film without knowing exactly why. Judging from the amount of flickering lighting I have seen so far I would suggest that there may be quite a few of these anomalies. Pay special attention to this because a film with many full frame rivalries throughout will repel many viewers.


Color rivalry occurs when unexpected lighting shifts cause the 2 eyes to have areas where the color values do not match. This is most likely in shots with large action or camera movements. This happens when a light bounces off a reflective surface directly into the lens. It will happen only in one eye if the I/O slider moves the camera enough to bring the lens out of the direct path of the reflection in question. The result will be that the reflecting surface will be much hotter and higher contrast in the corresponding eye. This is virtually impossible to prevent. Shots with extreme movement, particularly where the characters move very near to the lens should be examined carefully during blocking. If such an error occurs there are several techniques for repair. Reshooting should not be ruled out because some of these problems can take many days to fix.


1. In shots which are less round and deep the images in each eye will be substantially similar. By rotoing the corresponding areas in the good eye, these areas can be composited directly on top of the bad areas in the second eye, often with little distortion necessary. This will probably need to be done piecemeal because (for example) an entire face will have the features in slightly different positions and the perspective of each will be slightly different. The geometry of the bad eye can be used as an exact guide for getting this correct.

2. Color correction within articulate roto applied to the good eye. Since the problem is brightness & contrast based the good eye is the one which must be adjusted. This can be undesirable for aesthetic reasons, but doing the opposite is close to impossible in all but the most subtle cases.

3. In the most extreme cases, the corresponding features in the good eye will need to be warped, piece by piece, to fit the corresponding features in the bad eye. This requires a large amount of very intricate articulate roto. This is an extremely painstaking process which requires a high level of skill and patience. Perfect results can be achieved, but often taking much more time than it would take to reshoot the offending shot.

Leave a Reply

Your email address will not be published. Required fields are marked *