Previz and Blocking for G-Speak
Some shots are very, very hard to get right. How many takes must there be of the long studio shot in Magnolia? How many watches had to be synchronized to get the timing right for the pan across the ceiling-less apartment floor in Minority Report? But getting the shot, as tough as it is, is only a fraction of the difficulty: every hard shot is even harder to plan. How quickly is the camera moving? When should each actor hit their marks? Where are their marks, anyway? And what does the set look like? Does a wall have to be torn down? Is it starting to rain? How much are we paying these people, anyway?
To try and avoid these awkward questions—especially that last one—modern moviemakers use a family of techniques called “previsualization”, or “previz” for short. Previz uses tools like storyboarding, Matchbox cars, action figures, and 3D animation software to help a director, a cinematographer, a set designer, and other parties figure out what they want before paying a lot of money for actors and caterers to stand out in the rain. Most complex previz is done in programs like Autodesk Maya, which has a learning curve thatʼs basically vertical. So, the producer hires firms full of previz specialists to work with moviemakers to make representative 3D animations and environments. Unfortunately, if the cinematographer decides that the camera needs to be moved two inches to the left, she canʼt change it herself—it goes form her to the previz folks down the hall (or, in the worst case, over a cell phone), they work on it for a while, and she sees it the next day. By then, her feeling may be different, or she might have been misunderstood, and it takes another day to fix it. But she just wants to see what it would look like two inches to the left!
Crystal is an attempt to solve this problem. Using our most expressive input devices—the human hands—moviemakers can create a set on a tabletop, stage action-figure actors and Matchbox cars, frame shots with their hands, and view the result in a live 3D render. The movements of the camera and of the objects in the scene are recorded faithfully by the G-speak systemʼs motion tracker, they can be modified and played back, several takes can be recorded, and a shot can be planned with the ease of playing in a sandbox. The movements of cameras over time can be displayed to show coverage and set requirements, and the rough, easy-to- change results can be sent off to those professional previz folks for refinement.