The Rencontres Mondiales du Logiciel Libre is one of the largest Free Software events in the French-speaking world. This year, the event was held in Geneva, Switzerland, with the European Broadcasting Union co-hosting the Media, Radio, Television and Professional Graphics track of the conference. More than twenty projects related to audio-visual content production and distribution were featured in the track, covering ingest, encoding, storage, audio post-production, video-post production, graphics, play-out, and radio. A full run-down is available at http://schedule2012.rmll.info/-Media-radio-television-et-graphisme-professionnel-
Apertus (http://www.apertus.org) is working on Axiom, a crowd-funded open hardware camera capable of 4k resolution. Currently a working prototype, it offers an open alternative to the Red camera. It will not be as cheap as consumer DSLRs that can record HD video, but it will not have their limitations. It will be able to use various SLR lenses via a universal mount system.
SDI hardware cards are used in broadcast servers to ingest uncompressed video from cameras. They can cost up to 10,000 euro each, but typically the parts on the card only cost 500 euro. This is because broadcasters are perceived as being able to afford them. However, the drivers are often buggy, and vendors go out of business, leaving the broadcasters without support. A project was proposed by Darren Starr to create an open hardware SDI card.
OpenStack (http://www.openstack.org/) is a cloud platform being trialled by the BBC, versus proprietary cloud platforms. It can be installed on Ubuntu, but setup is complex and not consistently documented.
RAI, Italy's national broadcaster, has developed a media aggregator and archive system using open source tools (ANTS & Hyper Media News) which pulls media from the web, as well as live TV, and their own legacy archives in various formats, and segments streams into items. Then it uses natural language processing to analyze speech content and create searchable metadata from it. RAI has over one million hours of audio from its radio stations in its archive, plus several hundred thousand hours of video from its TV channels.
CasparCG (http://www.casparcg.com/) is used to playout graphics into news TV, including video overlays and huge LED-powered backdrops. It's used 24/7 by SVT, a national broadcaster in Sweden.
FreeLCS (http://freelcs.sourceforge.net/) is an EBU R128 batch processing tool which outputs loudness corrected WAVs and could be useful for post-production, helping end users record and mix at the right level. This level adjustment is traditionally done with meters, and a good producer will be consistent with themself, but it's hard to get consistent results between different producers without some kind of corrective tool.
In my presentation, I talked about the difference between traditional radio automation, with its heavy rotation music playlists, and social, crowd-sourced, human-curated broadcasting as enabled by Airtime. Radio stations can use Airtime to manage both traditional programme making and user generated content. Attempting to guess the audience's musical preferences via plugging and playlisting is over, now that the web has access to our individual purchase, search and recommendation histories. Furthermore, any MP3 player can shuffle and repeat a static playlist of current hits, making that particular radio format completely redundant. So radio automation may be dead, but radio can have a future as long as the humans are put back in charge of selecting and sequencing the content.
In the Open Source Broadcast Encoding presentation, the point was made that broadcasters use open standards such as SMPTE, so encoding from uncompressed to compressed formats is very feasible using open source tools. CPUs get more powerful all the time, so it is practical to replace an expensive FPGA based hardware encoder (which is probably running embedded GNU/Linux anyway) with a software encoder running on commodity server hardware. In response to concerns about increased power consumption from the server-based encoder, it was pointed out that dedicated hardware encoders can also drain a lot of power.
MXF (Material eXchange Format) is a metadata-rich format for exchanging multimedia projects. It is now possible to stream MXF with low latency using VLC, instead of an expensive encoder/decoder box, thanks to the work of Jérémie Rossier. PJSIP is an open source library which is used for two-way streaming in both voice over IP and broadcast audio contribution over IP (EBU N/ACIP) products. These are used for connecting remote studios, e.g. for interviews. In fact, EBU tests on the data streamed by 'black box' N/ACIP products revealed that they actually run PJSIP inside.
Interactive 'red button' television in Italy does not rely on having a broadband connection to the set, because many homes in the country do not have broadband. Instead, interactive data is interleaved with the digital TV signal. Avalpa has developed an open source datacast server for the DSMCC standard.
OpenHeadend (http://openheadend.tv/) systems can replace and combine satellite or terrestrial reception platforms, automation equipment, and digital recorders. A company has been set up in Paris to support the open source project. Thomas Kernen had built a fullly IP based TV distribution system 'head end' using open source, with a server featuring multiple tuner cards. All available channels from DTV and satellite sources were tuned to simultaneously, then forwarded to individual devices as required. The importance of a good quality motherboard was stressed, since the solution required multiple PCI Express cards which require full power to be available on the bus at all times.
Adaptive streaming is known by various names, depending on whether you're using Microsoft, Apple or Adobe versions. H.264 video files are encoded at various resolutions and qualities. The client device chooses the appropriate stream from a table, and if available bandwidth drops, switches to a lower bitrate stream. This type of stream can be accomplished using the open source x264 encoder. Apache modules for adaptive streaming are available (http://www.unified-streaming.com/) but the developers moved from open source to semi-open licensing. You can get the source only if you're a paying customer. Their earlier library was very widely used on commercial video sites, but there were no community contributions and very little demand for paid support, so they needed to find a new business model.
BBC R&D presented the results of the Limo project, part of an EU-funded large scale project on P2P stream delivery, including live P2P streaming. Apparently bandwidth is still an issue for the BBC, and multicast isn't happening on the public Internet, hence the trial of a P2P system. The Limo solution used HTML5 for interactive features and WebM for the video. It could be very interesting for Airtime as a more scalable, affordable alternative to Icecast unicast streaming.
During the Airtime workshop, methods for enabling video support were discussed. For DVB-T broadcast, it would be necessary to support hardware video output via something like an SDI card, or send video over IP to a broadcast device. This would be in addition to end-user distribution streaming (e.g. Ogg Theora stream via Icecast). Either Liquidsoap would need to be modified to support hardware video output, such as SDI, or Pypo would need to be extended to support playout other than via Liquidsoap. Of course there would also need to be some admin interface changes, such as allowing users to upload video files via the Add Media page.
OggStreamer (http://oggstreamer.wordpress.com/about/) is an open hardware streaming box which could be used to make it easier for Airtime contributors to hook up remote audio source streams. This product comes from the Otelo hacklab (http://www.otelo.or.at/) in Gmunden, Austria.
In the wrap-up session, broadcasters emphasized the need for open hardware in the specialist products that they rely on. The EBU offered to a host a similar event in future.