Ralph Keyser
August 3rd, 2007, 03:04 PM
The Duke City Shootout (www.dukecityshootout.com) happened here in Albuquerque last week. Now in it's 8th year, it's a film festival of sorts where 7 short films (12 minutes each) get shot, edited, and screened in 7 days. This year, the kind folks at Silicon Imaging lent us one of their prototypes of the SI-2K camera to shoot "The Spider Experiment". We promptly subjected it to four days of no-budget, seriously time constrained, independent filmmaking. I'll hope that I'm not jumping the gun on the SI guys by posting some thoughts and experiences from working with this camera. Any mistakes made here are my own. It's been more than a week since the end of production with several long nights during the post process, and my memory and notes may be missing in some spots.
"The Spider Experiment" is a charming little story about a 10 year-old girl trying to attract the attention of a boy in her class with an invented science experiment. I've always heard that you should never work with kids, animals, or new technology. We ended up doing all three on this project, and it actually worked out very well.
First off, this is a really different camera from anything I've worked with before. It's clear that the folks from SI did not start with any preconceived notions. From the touch-screen controls, to the computer standard connectors for I/O and video, to the raw data files plus metadata as output, this can not be confused with a video camera. I have to admit to some skepticism when we first started looking the camera over, but I quickly came to appreciate the elegance of the overall solution.
The biggest eye opener to me was the idea of recording the raw sensor data with very little processing done to it. The .look file you create in the camera is just a suggestion to the post folks, so you can push a lot of image look choices down the road. I can see this being a great thing since you can make those decisions in the ideal viewing environment of a post facility instead of an makeshift tent on set. On the other hand, as a cinematographer, I could see this as a huge loss of control over the final image, especially working under the traditional model where your involvement in post is minimal. If I was hired as a cinematographer on a project with this camera, I would want my deal to include being involved in some portion of the post process. This is really the same dilemma being faced by those working with the DI process today.
The camera itself is remarkably simple. There are none of the massive menus from a camera such as the F-900, and even though we didn't have a manual, it was always easy to find the function you wanted. Steve Nordhauser from Silicon Imaging spent about an hour giving us the guided tour, and we spent about an hour playing with the camera on the day before the shoot. I took a few notes on a 3x5 card, but after the first few hours of shooting I never referred back to those notes. The camera body only has a power switch and a cursor control cluster (sort of a stationary mouse) for times when you don't have the touch screen available. All of the other controls are in software. The work P+S Technik has done has resulted in the SI-2K having a very solid, well-built feel to it.
One astounding feature is the fact that the sensor head itself can be removed from the recording part of the body and cabled to the recorder or connected to a laptop running the camera software via a network cable. The head itself easily fits in the palm of your hand. Coupled with a small lens, you have a very nimble camera that opens up unusual shots and simplifies camera rigging. We had great fun getting shots such as inside the kitchen cabinet shots with no special work needed to accommodate the camera.
Because the camera stores raw data from the sensor, as long as you get the exposure and focus correct, the other things you set on the camera that effect the look are not applied in camera. There are several excellent tools available to help get the exposure and focus correct. For exposure, there is a pseudo-color mode that shows the percent luminance of each area in the scene, plus a histogram display. Since the lack of budget meant no waveform monitor on set, these tools were extremely useful, and I thought they were perhaps better tools than a waveform monitor anyway. For focus, there is a full-screen focus assist that removes the color and highlights the in-focus edges in red. It was one of the nicest focus assists that I've used.
Since this is a very limited budget festival, we took what we could get for glass on the camera. We were fortunate to latch onto a Canon 9x5.5mm HD zoom (B4 mount) along with some Fujinon C-mount lenses that Steve Nordhauser from Silicon Imaging was kind enough to provide. Changing between mounts is very easy with an adapter for PL or B4 mounts, and 4 allen screws let you add or remove the C mount.
Since we didn't need to go over 24 fps, we shot at the full resolution of the camera using the Cineform codec within .avi files since the Quicktime code is still being finished up. We used the Adobe CS3 suite for post with Cineform's Prospect 2K on a PC running Windows XP (we had difficulties getting Vista to run the post software correctly). We felt a little bit like the black sheep of the festival since everyone else was working in Final Cut, but we edited the full 2K Cineform files in real-time with no issues at all. I must say that the workflow was painless. We had an 80gig hard drive in the camera, and created a new folder for each scene. The camera automatically adds the date and time to the filename for each take. At lunch and at the end of the day, we pulled the hard drive and used a laptop to copy the files to a firewire drive that went to the editors. We never deleted anything off of the camera drive, and the drives are cheap enough that I could imaging just replacing them when they get full and keeping the camera drives until the shoot is over. The firewire drive went to the editors, who copied the files to the edit system. Once again, they didn't delete anything from the transfer drive. The .look files came in the folder with each scene, and they can be registered and applied to the footage with a couple of mouse clicks. As an aside, I was really pleased by the CS3 integration. It was very easy to go from Premiere to After Effects or Soundbooth for enhancing given clips. We used the tools in CS3 for the final color grading.
I know I've gone all this way without saying anything about the images from the camera. The short answer is that they were great. The material we were shooting did not push the limits on this camera, but I was impressed with the available dynamic range and overall flexibility of the images. The festival converted everything to DVCPRO HD at 720p for projection purposes, so I never saw a projection of the full resolution images, but by the time it got to the big screen, it still looked fantastic.
Of course, this was an early prototype and we were fighting the clock, so there were a few features that were missing, or that we didn't get to try out. We didn't, for example, have cables to get audio into the camera, and there are no XLR audio connectors on the camera. We shot double system to a Sound Devices 744t, which gave us very good results. We dumped audio data along with image data to the editor's hard drive. A little care with the slate and naming the files on the audio recorder made it easy to just sync sound as we went on the edit. We also did not have any way to synchronize timecode between the camera and sound recorder. My understanding is that there is an adapter, but the camera does not have any of the usual timecode connectors. This prototype also did not have a battery "fuel gauge", although I'm sure the production models will have one. As a result, we unexpectedly ran out of juice in the middle of a couple of shots, but we quickly got calibrated and started going by the built-in meters on the Anton Baur batteries.
I'm sure that this is not the camera for everyone. It clearly is aimed at the Digital Cinema market, and I would hesitate to chose it for applications that require fast turn around either in production or post. The camera is a special purpose laptop married to a sensor head. It takes time to go from totally off to ready to shoot. Not a lot, but it would probably drive you nuts in ENG or run-and-gun doc situations.
I'll have to apologize for not having any clips or frame grabs available from this shoot. The festival acts as the producer for all of these films and owns the rights to the images.
Overall, this looks like a very viable camera, and it's clear that Silicon Imaging is serious about being in this business. Many thanks to Steve Nordhauser for staying through the first day of shooting to get us off and running and to Jason for being our on-call safety net. We never had to use him, but in the middle of things on a Sunday, it was nice to know he was there. Special thanks to Silicon Imaging, Cineform, Adobe, and Intel for their support this year.
"The Spider Experiment" is a charming little story about a 10 year-old girl trying to attract the attention of a boy in her class with an invented science experiment. I've always heard that you should never work with kids, animals, or new technology. We ended up doing all three on this project, and it actually worked out very well.
First off, this is a really different camera from anything I've worked with before. It's clear that the folks from SI did not start with any preconceived notions. From the touch-screen controls, to the computer standard connectors for I/O and video, to the raw data files plus metadata as output, this can not be confused with a video camera. I have to admit to some skepticism when we first started looking the camera over, but I quickly came to appreciate the elegance of the overall solution.
The biggest eye opener to me was the idea of recording the raw sensor data with very little processing done to it. The .look file you create in the camera is just a suggestion to the post folks, so you can push a lot of image look choices down the road. I can see this being a great thing since you can make those decisions in the ideal viewing environment of a post facility instead of an makeshift tent on set. On the other hand, as a cinematographer, I could see this as a huge loss of control over the final image, especially working under the traditional model where your involvement in post is minimal. If I was hired as a cinematographer on a project with this camera, I would want my deal to include being involved in some portion of the post process. This is really the same dilemma being faced by those working with the DI process today.
The camera itself is remarkably simple. There are none of the massive menus from a camera such as the F-900, and even though we didn't have a manual, it was always easy to find the function you wanted. Steve Nordhauser from Silicon Imaging spent about an hour giving us the guided tour, and we spent about an hour playing with the camera on the day before the shoot. I took a few notes on a 3x5 card, but after the first few hours of shooting I never referred back to those notes. The camera body only has a power switch and a cursor control cluster (sort of a stationary mouse) for times when you don't have the touch screen available. All of the other controls are in software. The work P+S Technik has done has resulted in the SI-2K having a very solid, well-built feel to it.
One astounding feature is the fact that the sensor head itself can be removed from the recording part of the body and cabled to the recorder or connected to a laptop running the camera software via a network cable. The head itself easily fits in the palm of your hand. Coupled with a small lens, you have a very nimble camera that opens up unusual shots and simplifies camera rigging. We had great fun getting shots such as inside the kitchen cabinet shots with no special work needed to accommodate the camera.
Because the camera stores raw data from the sensor, as long as you get the exposure and focus correct, the other things you set on the camera that effect the look are not applied in camera. There are several excellent tools available to help get the exposure and focus correct. For exposure, there is a pseudo-color mode that shows the percent luminance of each area in the scene, plus a histogram display. Since the lack of budget meant no waveform monitor on set, these tools were extremely useful, and I thought they were perhaps better tools than a waveform monitor anyway. For focus, there is a full-screen focus assist that removes the color and highlights the in-focus edges in red. It was one of the nicest focus assists that I've used.
Since this is a very limited budget festival, we took what we could get for glass on the camera. We were fortunate to latch onto a Canon 9x5.5mm HD zoom (B4 mount) along with some Fujinon C-mount lenses that Steve Nordhauser from Silicon Imaging was kind enough to provide. Changing between mounts is very easy with an adapter for PL or B4 mounts, and 4 allen screws let you add or remove the C mount.
Since we didn't need to go over 24 fps, we shot at the full resolution of the camera using the Cineform codec within .avi files since the Quicktime code is still being finished up. We used the Adobe CS3 suite for post with Cineform's Prospect 2K on a PC running Windows XP (we had difficulties getting Vista to run the post software correctly). We felt a little bit like the black sheep of the festival since everyone else was working in Final Cut, but we edited the full 2K Cineform files in real-time with no issues at all. I must say that the workflow was painless. We had an 80gig hard drive in the camera, and created a new folder for each scene. The camera automatically adds the date and time to the filename for each take. At lunch and at the end of the day, we pulled the hard drive and used a laptop to copy the files to a firewire drive that went to the editors. We never deleted anything off of the camera drive, and the drives are cheap enough that I could imaging just replacing them when they get full and keeping the camera drives until the shoot is over. The firewire drive went to the editors, who copied the files to the edit system. Once again, they didn't delete anything from the transfer drive. The .look files came in the folder with each scene, and they can be registered and applied to the footage with a couple of mouse clicks. As an aside, I was really pleased by the CS3 integration. It was very easy to go from Premiere to After Effects or Soundbooth for enhancing given clips. We used the tools in CS3 for the final color grading.
I know I've gone all this way without saying anything about the images from the camera. The short answer is that they were great. The material we were shooting did not push the limits on this camera, but I was impressed with the available dynamic range and overall flexibility of the images. The festival converted everything to DVCPRO HD at 720p for projection purposes, so I never saw a projection of the full resolution images, but by the time it got to the big screen, it still looked fantastic.
Of course, this was an early prototype and we were fighting the clock, so there were a few features that were missing, or that we didn't get to try out. We didn't, for example, have cables to get audio into the camera, and there are no XLR audio connectors on the camera. We shot double system to a Sound Devices 744t, which gave us very good results. We dumped audio data along with image data to the editor's hard drive. A little care with the slate and naming the files on the audio recorder made it easy to just sync sound as we went on the edit. We also did not have any way to synchronize timecode between the camera and sound recorder. My understanding is that there is an adapter, but the camera does not have any of the usual timecode connectors. This prototype also did not have a battery "fuel gauge", although I'm sure the production models will have one. As a result, we unexpectedly ran out of juice in the middle of a couple of shots, but we quickly got calibrated and started going by the built-in meters on the Anton Baur batteries.
I'm sure that this is not the camera for everyone. It clearly is aimed at the Digital Cinema market, and I would hesitate to chose it for applications that require fast turn around either in production or post. The camera is a special purpose laptop married to a sensor head. It takes time to go from totally off to ready to shoot. Not a lot, but it would probably drive you nuts in ENG or run-and-gun doc situations.
I'll have to apologize for not having any clips or frame grabs available from this shoot. The festival acts as the producer for all of these films and owns the rights to the images.
Overall, this looks like a very viable camera, and it's clear that Silicon Imaging is serious about being in this business. Many thanks to Steve Nordhauser for staying through the first day of shooting to get us off and running and to Jason for being our on-call safety net. We never had to use him, but in the middle of things on a Sunday, it was nice to know he was there. Special thanks to Silicon Imaging, Cineform, Adobe, and Intel for their support this year.