The digitization of cinema opens up completely new possibilities for putting films on the screen – options once attainable only at a high price and with a great deal of effort. While the costs of digitization are largely borne by the cinemas, the related cost savings are accruing mainly to the benefit of programme suppliers. Apart from the major film industry companies and their distributors, these suppliers also include independent filmmakers, who can now make and distribute prints of their work much more easily than in the old celluloid days.
How this is done in practical terms and what has to be considered in the process are the subjects we will address in the following article. But first, a warning: this article is a complete departure from what has appeared on shortfilm.de to date. Namely, we will be explaining technical aspects of our theme for a change. Prompting this decision are the far-reaching changes that have occurred in recent years in the distribution and projection of films, and in particular the diverse, and in some cases distressing, experiences with digital films at the cinema. We will proceed, so to speak, from the top down, following the technical workflow from the most elaborate variation of D-Cinema to interesting alternatives in E-Cinema.
It all starts with a Digital Source Master
The first step after completion of a digital film is to make a master – the source material from which prints can be made in various variations. In D-Cinema this original version is known as the Digital Source Master (DSM).
DSM is not a standard, but rather a name describing original source material that is finished as a film but not yet designated or suitable for screening. Usually this is a digital film that was saved from an editing program onto a hard disk.
In-between step: Distribution Master
The next step in the workflow is a Digital Cinema Distribution Master (DCDM). The DCDM is the original print template in the highest possible quality in compressed or loss-free compressed form. Making a DCDM has the advantage that prints can be produced from it at the highest possible quality level for all later digital distribution channels, film-server types and playback formats. The disadvantage of an in-between step like this is that a DCDM places extremely high demands on processing capacity (computer performance) and requires a great deal of disk space. In many typical production processes the DCDM therefore does not exist as complete film file, but rather only virtually as a temporary frame-by-frame made during the production of a Digital Cinema Package and stored in the computer’s cache.
The final goal is to make a Digital Cinema Package (DCP). The DCP is the digital screening print for cinemas, festivals and other venues. It is usually delivered on a hard disk, from whence it is copied onto a local film server or media player. Unlike its predecessors in the workflow, in a DCP the film is in compressed form.
A DCP is a “˜package’ of digital films on a hard disk. There are norms regulating the way in which this package is put together. In 2005 the Digital Cinema Initiatives (DCI) in the USA, together with the SMPTE (Society of Motion Picture and Television Engineers), put together a set of specifications for digital cinema in 4K resolution, which it began propagating with the support of the major companies. This standard is in the meantime widely accepted. Apart from regulating data encryption, which was a prime motive for standardization, a DCI/SMPTE-compliant Digital Cinema Package has both mandatory and open attributes.
In technical terms, a Digital Cinema Package is a group of data files that store audio and video data separately, as well as the information needed to link them, so-called “˜descriptors’. While audio as a rule is not compressed, or is compressed loss-free (e.g. linear PCM 24 bit, 48kHz), the video elements must be stored and compressed in JPEG 2000 format in keeping with the DCI norm. JPEG 2000 is a codec with the special property that the film is not stored as a stream or in image lines (fields), but rather as a sequence of autonomous individual images, each of which are separately compressed – as in a JPEG photo. The container format for the individual JPEG images is MXF.
According to the DCI norm, each image must have 4K, i.e. 4000 horizontal pixels. According to the aspect ratio, this yields 4096 vertical x 1716 horizontal pixel lines for the classic cinema formats (scope) or 3996 x 2160 pixels (widescreen).
Producing a DCP
Theoretically, it is possible to make a DCP on a home PC. But the demands placed on hardware and software are high. And naturally special knowledge of encoding and conversion is necessary. In making a DCP much more has to be taken into account than only the above-mentioned parameters for container, codec, bitrate and image resolution. For example, the colour-space and audio-channel settings must also be programmed, as well as the display of subtitles.
DCPs can be created using special hardware or software. The hardware is however far too expensive for non-commercial, individual users. Software is more affordable, but more time-consuming to use. And there are not many applications available on the market. In Germany, the Fraunhofer Institute for Integrated Circuits – inventor of MPEG Audio Layer 2 (mp3) – has developed the application easyDCP. Marketed with the motto “as easy as 1 – 2 – 3″, the software is available for Mac or PC and is relatively reasonable. The simple version without encryption and player costs about 2,500 euros. And the prices are not likely to go down in future, as they always include the licence fee for the high-value JPEG 2000 codec included.
By now there are some open source developments available, but all are still in the beta phase. The best known is Open DCP (http://code.google.com/p/opendcp/), software that works with a licence-free JPG codec.
With this background in mind, filmmakers who release a short film only on occasion (and who are not computer nerds) would probably be best advised to refrain for the time being from embarking on the adventure of making a 4K DCP. Another aspect is that hardly anyone has a way of testing self-produced DCPs on film servers, of which in turn there are different types.
… perhaps a bit less would do?
Fewer demands are placed on hardware and software by the production of a DCP with storage in 2K resolution. This solution makes sense especially in cases when the source material itself is not in higher resolution anyway. A typical example would be a resolution of 1920 x 1080 pixels (with an aspect ratio of 16:9). ‘2K’ means that the number of horizontal pixels is about 2000. With respect to the widescreen format, the resolution would then be 1998 x 1080 pixels. For the scope aspect ratio, there would be 2048 x 858 pixels. The latter might pose a problem for the class of film servers that play DCPs up to 2000 pixels (see also the section on cinema equipment below).
However, DCI-compliant films in 2K resolution can be screened on almost all 4K systems in good quality.
DCP – good to know…
- Frame rate
DCI-compliant films are played at 24 fps (frames per second). This is called “˜true 24fps’, indicating a circumstance that many tend to overlook: only cinema cameras really record exactly 24 frames per second. Electronic cameras diverge from this rate, because at the time of transition from black-and-white to colour television one image line was sacrificed for control signals. The setting 24 fps thus actually means 23,976 frames per second. The same goes for 25 fps (or 30 fps in countries with 60 HZ), which is the actual standard for electronic cameras and all digital displays. The difference does not cause any problems for DCI-compliant screening, as in this case the films are transformed into a chain of discrete single images, but it does have implications when the DCI norm was not observed throughout post-production and until the moment of screening.
- Hard disks
Digital Cinema Packages are copied onto hard disks that are sent to cinemas. There, the data are copied onto the hard disk of the film server. It is important here that the film server is able to read the file system used on the hard disk supplying the data. What film servers require may differ from manufacturer to manufacturer. Films on commonly available hard disks formatted with Mac OS Extended/HFS or Windows FAT32 may cause problems, or be completely unreadable. File transfer usually functions smoothly however with the Windows PC Standard NTFS. Although Apple computers, which are widespread amongst filmmakers, are unable to save files in this format, there is a workaround available (see below).
- Disk space requirements
A ten-minute short film in 2K requires about 8 gigabytes disk space, although this is not a fixed size but depends, among other things, on the complexity of the images. An animated film with many monochromatic areas would be much smaller, for example.
Consider cinema and festival cinema equipment!
In most countries, DCI-compliant 4K digital cinema systems can be found only in big multiplex theatres. Most of the smaller cinemas that are of particular interest for short films cannot offer expensive facilities of this kind. It’s also important to remember that short film festivals usually take place in smaller cinemas and not multiplex theatres.
This is why it makes little sense to tailor a short film exclusively to the DCI standard. The next best quality option, which is available in many, though not all, small cinemas, is the use of 2K digital cinema systems. A different class of film servers is deployed here, which in turn call for a divergent set of specifications for the production of a DCP. It is not important here that the films do not then comply with the DCI norm because they are not digitally encrypted. More significant is that the DCPs have a lower image resolution of up to 2000 pixels.
In a few tests running DCPs by the Kurzfilmagentur Hamburg, it turned out that on a certain server model, films with 2048 x 1080 pixels could not be played, while those with a 16:9 aspect ratio and 1998 x 1080 pixels produced according to the same specifications (container, codec, bitrate etc.) could be screened trouble-free.
The reason is that there are different 2K DCP versions that use varying data containers: Full Container (2048 x 1080 pixels), Wide Container (2048 x 858 pixels) and Flat Container (1998 x 1080 pixels). The latter two formats (scope and widescreen) define at their maximum width and height the Full Container format.
Alternative distribution formats: High Definition
Most widespread and at the same time equipped to handle the greatest variety of standards are digital media players in combination with projectors or monitors with High Definition resolution. With today’s widely used 16:9 aspect ratio, High Definition would be equivalent to a resolution of 1920 x 1080 pixels. This resolution is supported, among others, by the following media players and distribution media:
- 1.9K (and higher) video projectors
- most of the new flat screens
- most of the new computer monitors
- so-called multi media players or media centers
- Blu-ray disk
- many video game consoles
- television in some countries (HDTV)
These players, monitors or distribution media are all downward-compatible, i.e. they also play films in standard resolution (SD).
If a Digital Source Master or Digital Cinema Distribution Master exists, prints can be adapted to these HD players or distribution channels (“down-scaling”). In this case as well, the digital films consist of data files with audio and video stored on a hard disk. However, there are no uniform norms for the production and formatting of films in High-Definition quality. Instead, there is an almost impenetrable variety of container formats and codecs, which in part depend on the computer operating system and the film-editing applications used.
Info: Container formats and codecs (incomplete survey)
First, it’s important to distinguish between container format and codec. A container format is, so to speak, only the wrapping for the video, audio and meta data. The container format defines the file structure and the instructions for how the files are to be linked. Some containers can also incorporate subtitles, others can’t. The container format can be identified by the film ending, e.g. <.avi>, <.mov> or <.mp4>/<.m4v>. A container format may include files that have been processed using various codecs. However, not all container formats can be combined with all codecs.
Codecs are algorithms that define both the encoding as well as the decoding of the audio and video streams. Codecs serve primarily to compress data, i.e. to reduce the disk space required to store them and also the required processing power. Each codec uses a different method to compress the signals. This method is what determines the audio and video quality. Relevant for film processing is also that different codecs take different times to process and that many codecs are suitable only for playback and not for editing. Frequently used video codecs are for example H.264 and MPEG-4. Frequently used audio formats are AAC, mp3 and LPCM.
Tips and remarks on containers and codecs
Not all end devices support all container formats. The same goes for (software) media players. The Quicktime player is for instance unable to read FLV, MKV or AVI containers. Conversely, the Quicktime container <.mov> is not compatible with many other media players.
For the home computer there are many useful plug-ins and drivers that help to overcome such hurdles. When a film is to be played on a different platform, one cannot simply assume that the necessary plug-ins will be installed there.
It is therefore essential to find out ahead of time which container formats and codecs are accepted by the system at the venue in question.
One way to be on the safe side, although not completely fail-safe, is to use containers and codes that are widespread and supported by many different devices. Not advisable are the more exotic formats such as DivX, Oggmedia or Matroska.
Note: The open source container Matroska <.mkv> is enjoying increasing popularity amongst filmmakers who work on PCs, where it is replacing the outdated AVI as video format for Windows. With Matroska it is possible to do things like create chapter titles and several sets of subtitles. It also supports the audio codec AC3 DTS (Dolby multichannel sound). The great flexibility of <.mkv> is at the same time its weakness, at least as long as there is no standard norm and no widely available hardware support for it besides PCs.
The most pragmatic combination at the moment, which is available to the largest number of potential viewers, is the use of codec H.264, the MPEG container <.mp4> and uncompressed audio (e.g. Linear PCM 48khz).
Codec H.264, also known as AVC, is a norm based on MPEG-4. It is actually MPEG-4 Part 10, of which also an open source variant exists (x264). This codec compresses data in good quality but it does require more processing power than some other codecs. H.264 is also compatible with AVI, 3GP and MKV containers, among others. The codec is used by multimedia players and Blu-ray and is implemented in AVCHD cameras and elsewhere. H.264 is also supported by the Windows Media Player and by Apple with Quicktime Version 7 and higher. Only on video consoles and “i-mobile” players might problems be encountered.
Note: The mp4 container is based on the Apple Quicktime file format and should be playable on any platform. On Apple computers, the nearly identical m4v container format exhibits a peculiar feature: after double-clicking on an m4v, iTunes opens, while an mp4 file starts the Quicktime player. The difference is that m4v was designed for Apple copy protection and does not recognize the AC3 audio format. Anyone who does not wish iTunes to be activated (which involves the automatic copying of data into the Media Library) can rename the files as mp4 without risk, or select the m4v file in the Quicktime player’s file-opening dialogue.
Another factor is the data rate. This determines the amount of video data processed per second during playback. If it is too slow, artifacts will appear; if too high, there is a risk that the hardware can’t keep up and the film might stutter during playback. In order to achieve optimal quality, the highest data rate should be chosen that is supported by the target medium and with which the file size remains within its capacity.
To give a reference: the maximum data rate for Blu-ray players is 48 MB/s. As a rule, though, a data rate between 15 and 30 MB/s should be sufficient to obtain good results.
However, the optimal data rate also depends on whether the video is interlaced or non-interlaced (progressive scan). If a Digital Distribution Master has been made, this tricky topic does not need to be addressed, for then the frames of the film are stored as single images.
Otherwise, most of the formats with the highest resolution consist of interlaced images, which means that each frame is divided into two fields of which the even and uneven horizontal lines are scanned respectively. This is the case with HD cameras and tube TV sets.
The International Telecommunication Union defines the highest HDTV standard as 1080i (1920 x 1080 pixels, interlaced), but the HD standard recommended for the time being is 720p (1280 x 720 pixels, progressive scan). With the higher standard, 50 or 60 half-frames (for 50 or 60 HZ) per second are scanned. The European Broadcasting Union by contrast recommends 1080p as future standard.
All new digital players and displays – such as DCI-compliant digital cinema – work with the progressive scan system. Many devices can in fact scan interlaced images and play them correctly, but the target format should be progressive-scan just in case. This makes it possible to make sure before distributing a film that it does not contain any image defects, because typically, when a film is converted from interlaced to progressive, errors occur when there are major movements in the image (“jitter”, or horizontal line patterns).
Multimedia players as low-budget alternative to film servers
Multimedia players or media centers are the contemporary successor to the video recorder as digital playback devices. With the right connections (HDMI) they are able to deliver videos up to a resolution of 1920 x 1080 pixels to monitors or video projectors. There are multimedia players with built-in hard disk or others that function only as a control unit for transmitting signals to an external hard disk. They are ideally suited for playing digital short films not only on flat screens, but also in high definition at the cinema.
Filmmakers can bring their own multimedia player with hard disk to the cinema with their film or a complete short-film programme stored on it, or send an external hard disk to a venue that has its own multimedia player.
Multimedia players with internal hard disk (“multimedia servers”) have the advantage that the image signals do not have to then be transmitted further (usually via USB cable – but firewire or eSata are faster), but rather can be streamed directly from the player. However, the hard disk should be large enough (recommended is at least 750 GB). The internal hard disk should be formatted for NFTS in order to be able to store longer short films “˜in one piece’, i.e. it should not have FAT32 formatting, for which the maximum file size is 4 GB.
When using Apple computers and periphery, an HFS-formatting of the hard disk in accordance with the Apple norm is possible, but only expedient if one is moving exclusively within the “˜iUniverse’. Otherwise, NFTS is the better choice even for Apple computers. Some functions – such as administration via USB with proprietary applications – are then no longer possible. But access to the NFTS multimedia player can still be enabled with the help of software add-ons (like Paragon, Tuxera or MacFuse + NTFS-3G; or using undocumented features in OS 10.6). The film data can then be copied via network cable rather than USB cable by mounting the multimedia player like an external drive.
The following additional criteria should also be heeded when choosing a suitable multimedia player: the connections and ports, the container formats and codecs supported, the functionality of the internal Graphic User Interface and the options for data management. As far as the containers and codecs are concerned, the rule is: as many as possible! There are already players under €200 that support all of the above-named formats, including the more exotic containers, DVD formats (VOB) and even pure camera data (e.g. in M2TS format).
Remarks and tips on the use of multimedia players at the cinema
Most media players do not yet offer any, or only insufficient, options for putting together video playlists. This makes it virtually impossible to show a programme of short films without interruption. Some multimedia players display navigation menus or icons on the screen shortly after a film starts. A way to circumvent both problems is to have an HDMI splitter/switcher and a preview monitor in the projection room. To prevent the problem of performing navigation steps in front of the viewing audience (as with DVDs), the films should begin with a few seconds of black or neutral screen before the first frame appears.
Another tip for projectionists: As most of the newer external hard drives, especially smaller pocket models, no longer have uniform professional ports – such as firewire/i.Link or eSata – projection rooms should be equipped with every conceivable type of USB adapter (USB A, USB B, USB A mini, USB micro etc.). This also applies incidentally to hard disks with Digital Cinema Packages – there have even been cases reported of films being delivered on a bare hard disk without housing and internal power supply unit!
General recommendations for programme planners, festivals and other short film event organizers
For film groups or film classes who would like to put on an evening of short films at the local cinema, one person should be put in charge of coordinating the screening with the cinema operator, including gathering together the individual films in a uniform standard and copying them all onto a single hard disk.
For short film festivals, it is important to clarify and communicate the specifications of the projection system to be used. Furthermore, the festival’s submission forms and regulations must be reviewed to ensure filmmakers know which formats are permissible. In the days of celluloid a few fields on the form specifying 16mm or 35mm were sufficient – with details like the aspect ratio identifiable on the film can. Virtually any cinema could then adapt on short notice, even with the naked eye by rolling the film. Today, a single checkbox for “˜digital’ is by no means adequate any more. For one thing, the accepted container formats and codecs must be listed as multiple-choice input fields. And the regulations should also indicate the acceptable storage media (e.g. hard disks, HD-DVDs) and their specifications.
Those putting together touring programmes or short-film reels as Digital Cinema Packages must weigh whether it is more practical to collect all films in a programme into a single continuous playback file or to copy each film individually onto the hard disk. If the films are to be played from film servers, the latter is advisable, as this is the only way to include short breaks between films – for example to make announcements or hold discussions. Even if no breaks are required, it is still relatively easy on professional film servers to put together a playlist of individual film files for a continuous programme.
In the case of festival retrospectives, this would make it possible to copy all films, even those not destined for screening at a particular venue, from a Digital Distribution Master. At least if the organizers at the venue can be trusted to show only what they have ordered and paid for 😉
Whether distributors, organizers or filmmakers, all are forced to come to terms today with digital formats and their internal parameters. But it is by all means worthwhile to acquire a good grounding in this field and to continue to keep up with the latest developments. It has never before been this easy to play films on every conceivable kind of display or screen, to adapt them to various platforms and to distribute them in high quality at low cost.
Appendix – Links
- Super (Windows)
- ffdshow (Windows)
- MPEG Streamclip (Mac)
- freie Codecs
- JES Deinterlacer (Mac)