Total cost other than air fares but including hotels, food, conference fee etc was £147.
The major reason for this visit was the annual UAIDE meeting which this year was held in Los Angeles. I flew out direct on the Sunday. The flight was 12 hours non-stop and was a trifle long.
The major difference between this and the last one I attended was the difference in size. The credit squeeze in the USA has seriously changed the kind of support that conferences are getting. At the animation sessions there were between 30 and 40 people this time compared with 150 two years ago. Even so, most of the people directly involved with animation to any large extent did attend. Some of the names:
On the animation side, probably the most impressive paper was by Francis Honey concerning CAESAR. I will discuss this later as part of the write up of the visit to Denver.
The next most interesting was Burtnyk and Wein on the National Research Council of Canada's Animation System. This is a system on a machine about the same size as our PDP15 with a larger disc backing store. The user builds up a library of key frames between which he interpolates. The system is similar to our own as far as facilities are concerned. However, it is oriented more towards the conventional animator. The smallness of the computer restricts them to a maximum of 15 different cels at any one time. One interesting point, also made by Francis Honey, was that there was a need to pick up an in-between frame and later redefine it as a key frame. The production process often started by defining a complete sequence by two key frames and interpolating between them. A frame somewhere in between may then be isolated and different timings for the in-betweens in the first and second halves adopted.
They have some simple scan conversion algorithms for filling in solids. Basically changing parity on every line intersection. Their playback on the display tends to be at about 1 frame per second. The scan conversion often produced good artistic effects.
Nestor Burtnyk showed two films. The first METADATA was really just in-betweening between random pictures and was not very impressive . A second cartoon about a man and a restaurant (La Faim) was quite impressive in the quality of the human movement. It also showed the limitations that they work under. Each frame was too large for their internal buffers as they can only hold a thousand points. Consequently they were having to do it by breaking the film up into bits and rewinding the film between displaying each bit. They have to do this 7 times in this example.
They have a white phosphor CRT tube and get their colour films by a programmed colour wheel inserted in front of the camera lens.
He finished by showing a third film giving sequences from a number of commercial TV broadcasts. All were of interest but not that special.
A session on Computer Art provided some interesting film but not directly relevant to us. Ken Knowlton described his TARPS language which was used to make the Stan Vanderbeek films. Jack Citron of IBM described some work they had been doing with an artist Jesse Reichek who paints pictures consisting of only two shapes in various positions. An analysis of his style was done and a computer system developed to allow him to make the kind of transformations he had used in his previous pictures. He was then able to use the system to define other pictures and also films consistent with his style.
There were a number of other animation papers which did not impress too much. However, there were two good papers on 3D animation. A Japanese system at Hitachi was described which used Coons's patches to do curve fitting, around real life objects such as a duck and a human body.
The other paper from the MAGI company was concerned with scenes made up from a small set of shapes such as cubes, ellipsoids, etc. The light source and camera position had to be defined so that pictures could be produced with correct light and dark shading. The results were impressive with highway-views and tanks being shown. However at a cost of 30 secs of CDC 6600 time per frame it was very expensive.
There was the usual UAIDE film show one evening. I was impressed by the photographic qualify of the films now being shown. Two years ago it was sufficient to have a good scientific problem requiring animation. Now it must also be in colour with a soundtrack and professionally finished off. In a 3 hour show there were only 2 or 3 films which fell below this standard. Listing some of the ones I remember:
PROTEIN PRIMER will soon be available from Harper and Row on a commercial basis. It has a voice soundtrack and labelling in addition to what we have in the draft copy in the library.
GENESYS can now be obtained from a number of photographic laboratories in the Boston area. Permission to buy a copy must first be obtained from Jack Mitchell, Digital Computers Group, MIT Lincoln Laboratory, Lexington Mass 02113. Cost is about $60.
Other titbits of animation information:
I attempted to find out answers to a number of problems we have without too much success. The first of these is the mounting of the new pin registered camera in the SD4020. Julie Landstein (IBM) did not know whether they could adjust their camera to give a full academy image. She suggests we contact either Bill Deorazio or Paul Grier at Yorktown Heights who may know. Talking to Paul Ressler of Datagraphix on the same subject, he thought that it was possible to get the full academy image. He did not think they made a variable mount and could see no objections to us installing one of our own. They had had some trouble with the 35mm action on the 4020. The fault appears as incorrect frame advancing. Apparently the pulse length on the 4020 is slightly shorter than on the 4060. If the camera is correctly lubricated the action should be correct. The solution is therefore to get the customer engineer to increase the pulse length. Also warming the movement up by some means can cure the fault temporarily. If it happens the important thing is to let the USA know.
The reconditioned SC4020 models for sale at the moment tend to be basic machines with no cameras. If any options are on them it is purely because they were on them before. They will not be added without us being charged. One of the reconditioned machines is being given to the animation side of UAIDE. The machine will be located at Maryland and will be available virtually free to anyone doing animation who has insufficient funds.
Julie Landstein did not know exactly the levels of intensity provided by the variable intensity option. However she felt that the darkest level was not very great; that is similar to our current situation. Again if we contact her engineers they would be able to give a precise answer.
A number of products, some new, were described by Datagraphix, listing these:
The main purpose of this visit was to see their new hybrid computer CAESAR (Computer Animated Episodes using Single Axis Rotation). Computer Image have been producing commercially acceptable animation for the last few years, first with ANIMATE, then SCANIMAC, and now CAESAR.
I saw the first two systems working and also a number of partly produced SCANIMAC systems. I think they now have three SCANIMAC systems working in New York, one in Los Angeles as well as the equipment at Denver. SCANIMAC allows you to do 3D effects on simple art work. It is primarily designed for titling and special effects. It is impressive but has limited applications. The system consists basically of art work, TV camera, hybrid computer, black and white video output. The colour output is done by optical superimposition afterwards.
CAESAR, on the other hand, is designed to produce commercially acceptable coloured video output directly. Suppose we are going to generate a cartoon of a rabbit running across the screen in front of a background of desert sand and cactii. (This was actually being done as I was there). Initially, the background has to be produced by an artist and placed in front of a video camera. They had a single background representing, a number of different scenes, and picked the desired one purely by moving the camera. There is no internal storage of the background. Consequently it has to be positioned accurately and left in place during the whole time the animation is taking place.
Next the artwork is required for the figure to be animated. In the case under consideration this consisted of a rabbit head, body, arms and legs on a single sheet. The positioning of the parts is irrelevant. The artwork is produced in five different grey levels. The rabbit I saw only had three, a light and dark grey, plus white. Again, there is no permanent internal storage of the artwork. The memory consists of the original drawings scanned by a video camera once more. In the set-up at the moment, the background and video camera are in a separate room from the figure to be animated and its associated camera.
The individual parts of the rabbit are now located and differentiated by hardware. From now on they can be thought of as rectangles containing an object in 5 grey levels. The artist now has controls to make the following adjustments. For each frame of a sequence he can:
The parameter information described above for a sequence is stored in a Honeywell DDP 316, the digital part of the system. The animator sits in front of this mass of controls manipulating the relevant functions, playing it back a frame at a time, making adjustments, recycling, etc. He can either display at real time speed or at 2 frames per second. Although there is a large number of controls, they are probably as few as possible. For example, three continuously variable controls give the amount of each primary colour. It is a simple matter to change the colour of the rabbit at will. While I was there the animation had been just about sorted out and they were aligning the background correctly, This seemed purely artistic. They were trying to decide how much of a rock should be visible.
Hidden line elimination is done on a basis of five levels. Usually the background is one level and the individual parts of the animation are assigned other values. This way one character can walk in front of another. In one scene that was played back to me, the background had been divided into three levels, two of which were parts of a stream, The scene showed a coyote floating in the water with variable parts of him being in view as he undulated along. It seems like a 20 second sequence might take a few hours to get right. About a third of this time is devoted, to sorting out the animation. The analogue functions sometimes seem to want to bend the arm in unusual positions! Over half is spent in the artist getting the colouring correct while the rest is probably sorting out technical problems.
Computer Image have a trained crew of artist, video man, controller, etc. who work with the user (the animation director) in producing the result.
Once completed the sequence is run off and stored on 2 inch video tape. Editing of this will eventually produce the final product.
The amount of artwork required by this technique is considerably less than by the conventional method. A whole scene can be produced from a single picture of a figure, The company have only had the equipment working for two months and during that time they have completed two films. The first is called QUARTETTE CAESAR and the second THE COYOTE AND HORNED TOAD. The first film consists of a number of sequences involving a cartoon character called CAESAR and really illustrates some of the facilities available. Only 6 drawings of CAESAR were required for a five minute film.
The second film is part of the Navajo project. This is a set of school films for Navajo Indians depicting stories which are legendary. In the first film the coyote comes upon a horned toad with a field of corn. He demands some corn and the horned toad gives it to him. Eventually the coyote gets more greedy, eats the whole field and the horned toad. The horned toad, however, does not die and gives the coyote a bad time inside him. The coyote tries various ways of killing him. For example, he drinks water, attempting to drown him, and so on. This gives some indication of the scope of the story. It was done with only 46 separate drawings.
Three Navajo Indians completed it while working with the Computer Image staff. Thus it was quite amateurish as far as the animation was concerned. Yet the quality of animation is as good if not better, than Walt Disney. One scene, showing the horned toad walking slowly across the screen with a limp after getting out of the coyote, is one which would very seldom be attempted by a conventional animator.
Computer Image charged $1000 per minute for the Navajo film, although it cost around $2250 per minute to produce. An estimated commercial price would be about $7000 - $8000 per minute. Currently Computer Image have two crews working 9 hours a day using the equipment. An artistic director works a day, split across the days of the technicians. They are currently working at about 10 minutes of animation per month. There is a possibility of them doing a Saturday morning series. This would, probably require 3 CAESARs.
The only criticism of the current hardware is that the video equipment is sufficiently complex that noise gets into the system appearing as white specks on the characters. It is necessary to tune the equipment up before generating the final output. The major problem while I was there seemed to be getting rid of some white spots on the rabbit.
Francis Honey thinks they can generate commercially acceptable film rather than video output, but it will require considerable additional equipment. He will wait until they have made a commercial success of what they have at the moment. The film generation would not then be real time.
There is really little one can say about Computer Image without using superlatives. Each year they have made tremendous strides. It seemed impossible that they could get this far in so short a time. They have been the only company interested in making hardware aimed specifically at the conventional animator, talking his language. The anthropological harness, for example, which allows a person to make the movements he wishes to be followed by a cartoon character was abandoned purely because it was not liked by the animator. It was not the way he did things.
Already Computer Image are thinking in terms of 3-dimensional animation. If they get sufficient support it could well happen next year or the year after.
The expertise and enthusiasm generated at Computer Image is infectious. They are a small bunch of dedicated professionals who really are making an impressive product.
I tried to get Francis Honey to let me have a copy of one of their two films for the library. He is now completely unhappy with the CAESAR film and does not want to show it around too much. The Navajo film may be available in about two months if we write to him. I have a short clip of some demonstration sequences they did. The quality is poor and he does not wish me to show it without making the relevant apologies.
I talked to Francis Honey about PMN's efforts to get colour output from our VT15 display. As he does not regard us as competitors in any way he would be prepared to manufacture for us a system which would take the display picture at 5 grey levels and transform it to a colour image with complete flexibility of colouring of the individual levels. He has virtually all the hardware now and could do it for about $10,000. The maintenance required would be practically zero. They have had five similar systems running for about a year with virtually no maintenance.
On paper the Moore School has probably the best integrated animation system in the USA. Unfortunately their published work does not reflect the actual situation. On the hardware side they have a SPECTRA 70/46 and linked to this over telephone lines is a DEC 338 and PDP8 system. The software basically consists of a number of theses:
Bernstein and Deily should be available through University Microfilms, High Wycombe.
Unfortunately they had no spare copies of any of their theses. I tried to see some of the films they had produced and, after spending about half the morning looking for a projector, was able to see the first film they had produced. This is a long one in colour on electromagnetic theory. The film is good (runs for about 30 - 40 minutes) and was produced at Brooklyn using optical work to get the colour. The jitter from the SC4020 camera was very apparent.
The second film on the same subject is under construction locally. They have a slave 338 tube with a greenish phosphor. An animation camera is placed in front of this. They have a small photoelectric cell which activates the camera from one of the PDP8 lights. The camera shutter opens for a finite length of time. During the time the shutter is open, the frame is displayed on the 338 tube 8 times. Thus they work in a way very similar to the SC4020. To get colour, they insert filters between the camera and tube. They produce all the output for one colour, back up the film, then do the next.
The film generation is done from DEC tape. They keep display files in a compressed form and are able to display frames faster than real time if desired (basically only keeping the difference between one frame and the next). The DEC system was down while I was there so I saw none of this working.
My impression is that they have no longer any real interest in animation which seems a complete waste.
One other film I saw was of a horse walking, trotting, etc., done by a Japanese. The film was quite impressive in the life-like motion. Unfortunately he wrote up his Masters in Japanese (never learned to speak English).
I am afraid this was a rather wasted day.
T visit was mainly to see Woody Anderson and find out what his latest animation work was. The Applied Physics Laboratory is quite a way from the main University and looks more like a Government laboratory. That is, it has armed guards; passes are required for all workers; visitors have to be supervised at all times, etc. Apparently it does a large amount of work for the Navy.
The main hardware is a 360/91 with 2 million bytes of core, a host of disk drives and input mainly through IBM alphanumeric displays situated in offices. Connected to this is an IBM 1130 with 32k of core and two disk drives plus usual peripherals. They have a 2250/4 interactive display attached to this. Anderson now has both CAMP and CAMPER systems resident in the 1130. He has an editor which allows him to correct the animations followed by an immediate run of the job. The turnround achieved in the stand-alone system was quite impressive. The speed of output was about 1 frame/second. Debugging film could easily be done quite quickly. Once a sequence was completed an actual SD4020 run was made in the 360/91 and the magnetic tape generated. Anderson has only room for 5 stacks in the 1130 instead of the usual 8. He has also removed a number of the infrequently used routines.
He is currently working on a number of film pieces. The most elaborate is a three screen production with Stan Vanderbeek. Sequences involving a number of 3-dimensional rotating shapes start on the middle screen and spill over into the two side screens as zoom-ins occur.
The 1130 system appears to have only two users so that Anderson can have virtually unlimited time in the stand-alone mode. He is also working on a number of other interactive systems. These include:
Currently his main work appears to be the generation of a number of applications packages for the 1130. His animation work is currently a sideline. He is hoping to get more people interested in the system but currently only has about one user.
Most of his SD4020 output is now done by Joseph Kaye & Co. in Boston. The charges made are:
The major purpose for visiting Boston was to see Sherrill Martin of Joseph. Kaye & Co. They have continued to provide a high quality service for the production of computer generated films. The company is either prepared to make the film right from the storyboard or alternatively will do the microfilm recording and optical work given the magnetic tapes.
Sherrill Martin has continued to use the SD4020 at MIT, the SD4060 at Lincoln Labs, the BETACOM 600 or CALCOMP 890. The last two are available at local computer bureaus, while the first two are available free to some University users.
He has continued to use his Flight Research 35 mm camera as the main method of output. He has a number of lenses for the camera depending on the recorder being used. The standard lens holder has been replaced by a PENTAX thread onto which he mounts the relevant lens via an adaptor. One particular lens that has proved useful is the NIKON CRT NIKKOR which is designed specifically to give good results with a P11 phosphor and a flat image field. It is an F1.2 lens and costs about $213. He uses this on the BETACOM.
Currently he prefers to use the CALCOMP 890. He feels that the 30 intensity levels give him enough to produce smooth fades and dissolves as long as they are not extended (ie must be less than 100 frames). In addition with the point rather than vector mode of drawing, he feels there is a much easier setting up situation. He has little trouble with variable vector intensity. Also the 3000 × 4000 image he uses gives adequate resolution for the production of smooth curves. The major drawback is that towards the edge, there is distortion. Consequently, although the screen size is 7.5 × 10, he uses only the middle 8 in the X-direction. He finds the external monitor particularly useful in as far as he can abort runs as soon as he notices trouble.
The maximum speed he has obtained on the 890 is about 8000 frames/hour. He finds the setting up of the machine particularly easy. There is a meter on it which gives the cathode current in the CRT. Keeping this constant has given him constant output for the last year. He feels the intensity control on the 890 is very impressive. The tape drive however is quite slow, especially in rewind. On simulation of SD4020 tapes it runs at about 4000 frames per hour. This compares with 6000 for the BETACOM and 15000 for the SD4020 itself.
He does not seem to have had anywhere near as much trouble as us in setting up the SD4020. He only alters the Short and Long Vector Controls plus the Intensity control. He finds the position of the Uppermost control (short vector) is usually near its maximum value, while the next (long vector) has a setting around 1.5 to 2. 5. The Beam setting control is around 3.1. Increasing the brightness control above a certain level makes it impossible to get a balance. He thinks that given adequate tube intensity it is a trivial matter to get long and short vector intensity the same. This however is not the case if there is inadequate intensity.
One problem he has had on the SD4020 which may worry us - the frame counter cannot keep up with the film mechanism if the camera is at all fast. Consequently, incorrect readings result. In addition, the afterglow on the tube is sufficiently great that quite frequently the next frame gets a ghost of the previous frame if the camera advance is fast. This does not worry cine output but could worry microfilm users.
One interesting point. Nelson Max has been using an advance repeat feature on the SD4060 at Lincoln Labs and has come up with identical problems as us. As soon as a tape block is misread he gets into trouble.
On all machines his standard method of operation is to produce a complete film on one run. Apart from tape errors he would expect no faults on the SD4020 over a 4 hour period. He has in fact run the SD4060 continuously as long as 11 hours on one run. Nelson Max, for example, generated his Peano curve film in two 11 hour runs using the equivalent of 60 magnetic tapes. Some frames took as long as 4 minutes to generate.
If odd frames are corrupted he will edit these out optically. He starts with a 35mm negative, makes a 35mm high contrast from this and produces the 16mm inter negative from this. He estimates that the complete cost including programming works out between $500 and $2000 per minute with most films in the $500 - $600 range.
Additional people talked to at MIT included Judah Schwartz, Wade Shaw, Bob Conrod and Jean Bow.
