Paul Cameron and Dion Beebe, ACS push hi-def video to its limits for Collateral, which chronicles a hit man’s nocturnal killing spree.

In the new film Collateral, a lonely taxi driver, Max (Jamie Foxx), agrees to chauffeur the smooth-talking Vincent (Tom Cruise) around Los Angeles for an entire evening. He soon discovers that Vincent is running an unusual errand indeed: he is a mercenary who is methodically eliminating five witnesses scheduled to testify against a drug cartel in federal court. Unable to escape Vincent’s grasp, Max quickly becomes the chief suspect in the murders, and as federal and local law-enforcement officials close in on the duo, Max realizes his only way out is to prevent Vincent’s final murder.

Collateral director Michael Mann had experimented with high-definition (HD) video for a few scenes in Ali (see AC Nov. ’01), and he went on to produce the television drama Robbery Homicide Division, which was shot entirely with Sony/Panavision 24p CineAlta HDW-F900 cameras (AC Feb. ’03). Intrigued by the format’s potential for feature filmmaking, Mann decided to use it on the extensive night-exterior work in Collateral to make the most of available light in and around Los Angeles. “Using HD was something Michael had already settled on by the time I came aboard,” recalls director of photography Paul Cameron (Man on Fire, Gone in 60 Seconds), who prepped Collateral and shot the first three weeks of principal photography. “He wanted to use the format to create a kind of glowing urban environment; the goal was to make the L.A. night as much of a character in the story as Vincent and Max were.”

Cameron spent several weeks testing the available 24p HD packages — the Thomson Grass Valley Viper FilmStream, the Sony/Panavision F900, and the Sony F950 — and comparing the 35mm film-outs to footage shot on four high-speed Kodak stocks: Vision 500T 5279, Vision2 500T 5218, Vision2 Expression 500T 5229 and Vision 800T 5289. (He pushed all of the stocks by two stops, rating the 500-speed emulsions at ISO 1600 and the 800-speed emulsion at ISO 2500.) “It became pretty apparent early on that the only way to get the look Michael wanted was to shoot HD and push the gain on the cameras,” says Cameron.

“We ruled out the Sony F950 early on, mostly because of the optics,” continues the cinematographer. “I felt the image was a bit too soft, even with Zeiss DigiPrime lenses. Another drawback to the F950 was that you had to have a separate recorder, which meant we would always be tethered to that deck with an umbilical cord. Although many improvements were made to that system on Star Wars: Episode III, I didn’t feel it was production-ready. The F900, on the other hand, is a tried-and-true workhorse, and it yielded absolutely sharp results with the DigiPrimes.”

The Thomson Viper has been lauded for its 4:4:4 uncompressed raw-data FilmStream mode, in which the pure image signal is sent to a hard drive, but Cameron found that this mode of shooting posed several practical problems. Because the signal undergoes no processing and is viewed on a monitor in “raw” form, the resultant images have a sickly, greenish hue. “That yellow-green image didn’t represent anything we could see by eye,” says Cameron, “and that made it impossible to judge the image from the monitor. If we passed the signal through an RGB processor and looked at a composite image, we were looking at a very compromised picture that didn’t reflect the image quality of the FilmStream mode at all. Also, when we increased the gain in the FilmStream mode, we weren’t seeing any results. In order to get the look Michael wanted, we had to push the camera beyond its normal mode.”

Cameron discovered that although shooting in VideoStream mode resulted in a slightly compressed image, the ability to modify the image via the camera’s internal matrix menus was a much better choice for the look that Mann wanted, and resulted in a superior picture for the film-out tests. Still, there were additional problems. “The Viper was just not production-ready,” says Cameron. “First off, we had to manufacture several accessories for the camera — a mattebox, new lens rods and base plates.” And because the Viper is not equipped with any recording device [internal deck or hard drive], “we had to make a fiber-optic umbilical to attach the Viper to the recording decks, and it’s difficult to work with any kind of tether.

“You can record into an HD deck or you can record to a hard drive,” explains Cameron. “We looked at the Director’s Friend/DVS CineControl [see New Products, AC Dec. ’03], but that device is far too large and cumbersome to lug around in real-world shooting. Bryan Carroll, Michael Mann’s technical adviser, came up with the S-2, a hard drive that’s about as big as a 1000-foot-magazine case. The system wasn’t specifically designed for the Viper, but it was applicable. However, it was difficult to retrieve the material from the hard drive, and that sent up a red flag with the studio; they demanded that we have a backup system if we decided to shoot with the Viper, so suddenly we were looking at the S-2 drives and an HD recorder. The gear attached to the camera was doubling.”

The S-2 drives had a recording capacity of about 40 minutes, but in the end, the filmmakers decided not to use the redundant hard drives. Instead, they opted to simply record the Viper VideoStream signal to the new Sony SRW5000 HD deck, a proprietary system manufactured for the Sony F950. The production had four prototype SRW5000 decks, two in the field with the Vipers and two at Laser Pacific, where the HD footage was transferred to 35mm.

The production began shooting with a pair of Vipers, but Cameron testifies that “by the end of the first day, we switched to two F900s as our main cameras.”

To achieve the light sensitivity Mann was after, Cameron found that he had to push the cameras’ image gain in order to achieve exposure in the available ambient light. He found an acceptable signal-to-noise tradeoff within the range of +3dB to +6dB on the F900. “There is a very thin line with the threshold of noise in all of the HD systems,” he explains. “In testing the F900, we found that some things looked great on the monitor at +3dB or +6dB, but when we made the film-out, those scenes picked up even more noise when projected on film on the big screen.

<< previous || next >>

© 2004 American Cinematographer.