UPDATE 10/23/2017: Several significant updates are coming soon, which may be worth waiting for if you're just startingout. If you're looking into this project, you may want to wait check back in a week or so, when I expect these to be available. These include:

• Port to Python 3 and Qt5 (currently uses python 2.7 and qt4)
• Redesigned UI to reduce need for frequent switching between panels
• Executables for linux, to avoid the complicated install/compilation process currently necessary (hopefully OSX & windows coming later)
• Flat-field correction and distortion correction to compensate for effects introduced by non-stock lenses
• Improved documentation

Update: If you're implementing this and want suggestions/advice from other users, consider joining the google group

Design:

• Film transport is via a repurposed projector (8mm, Super8, or 16mm), driven by a stepper motor.
• Raspberry Pi acts as a headless server, controlling the camera, driving the motor, and streaming captured images over a network to a more powerful client.
• Python 'client' program runs on a higher-power computer on the same network. It sends control commands to the Pi server, and performs processing on images as they are captured.

Optimizations for Speed:

• Streaming images over network reduces I/O delay between frames, compared to saving them locally.
• Multithreading on both client and server further reduces I/O delays.
• All computationally demanding processing (esp. image fusion; see below) is performed by the (faster) client.
• Can capture via the Pi Camera's (slower) stills port, or can blend captures from the video port for increased speed.
• Practical per-frame capture times range from 1.6s/frame for 3-exposure bracketing from still port, down to .5s/frame for single-exposure captures from video port.

Optimizations for Quality:

• Performs exposure bracketing on frames and combine them, to capture a much higher dynamic range than the Pi Camera normally permits.
• Full control over Pi Camera's settings, some of which may be changed mid-capture.
• Some simple histogram adjustment tools can do some processing on captured images before saving.

Optimizations for Usability:

• 'Setup Mode' allows for precise adjustment of camera before starting capture. <img align="right"src="images/pifilm-capture.png">
• Near-real-time view of captured images.
• Save/load settings, e.g. for different film types or projector setups.
• 'Smart capture' features to adjust to dramatic lighting changes (under development)
• Entire system can be moved from one projector to another to change film gauges.

See the wiki(still under construction) for complete documentation, including a manual, wiring diagrams, and hardware selection tips for your own project.

Demonstration video, samples, and comparison w/ direct capture at speed available on my YouTube channel. Many more capture examples on my Vimeo channel