Session 3: 3D Conversion
Moderator(s):
Brad Collar, Vice President, Technology, Warner Bros.
Panelist(s):
Barry Sandrew, Ph.D, Founder and President/COO, Legend3-D, A Legend Films Company
Chris Bond, President – View D, Prime Focus
Chris Yewdall, Executive Director and Chief Executive Officer, DDD USA, Inc.
Warren Littlefield, President, The Littlefield Company
Brad Collar
There are often two controversies when the entertainment industry looks at a new technology; should we do it, and what is the best way to do it? The list of options in 3D keeps growing. This panel will focus on 2D-to-3D conversion.
Warren Littlefield
Home 3D is the next breakthrough. In 1996, he did a 3D episode of 3rd Rock from the Sun. It doubled their audience, and helped them win the May sweep. The lesson learned was that the audience was looking for something new.
There are three categories of 3D content to the home; filmed entertainment, live content, and games.
Four business models (Warren’s term):
– Home video: over 600 episodes of StarTrek are available for a 3D sale. He thinks 3D conversion will be the killer app
– 3D networks – expect more this year
– Worldwide program distribution
– Digital distribution, including iTunes, Amazon, etc.
The number of sets that will be sold is unknown. With great cost comes great risk. The revenue models will depend on how many 3D sets are sold. There is no format war, just a need for content. This is an opportunity to up-sell the library. The box office for 3D indicates a sustained appetite, dispelling the idea of ‘stunt’ appeal. Building the market will be about volume – volume of set sales and volume of content (live events, movies, and series) together (the chicken plus the egg).
Chris Yewdall
(Chris calls himself the self-proclaimed minister of propaganda for 3D)
Conversion can be done at one of three quality levels: good (embedded in 3D device, automated 3D conversion), better (3D conversion with automated depth recover, manual focal point/depth effect decisions), and best (original 3D content creation, or 3D conversion with manual / semi automated depth recover). The cost goes up accordingly.
Carl Franklin wrote a great book, “Why Technology Fails.” Any 3D conversion must pass the “so what” test.
Since 1993 DDD has been developing a process called Depth-Image Based Rendering (DIBR). A single 2D image is depth mapped using monocular queues to create the conversion. They have been working to make the manual aspects more efficient by continually improving the computational rules. Cost went from $100k/minute in 1999 to $1,500/min in 2007 by relying on the automation more and the human intervention much less.
3D conversion consists of two stages: depth recovery from 2D content, and 3D scene reconstruction from depth and source image. Auto 3D conversion will not deliver the “better” category (first paragraph). Human intervention is required, particularly for 3D focal point. At the end of the day, the source content is still 2D, but filming in 2D with the scene set-ups designed for 3D improves the results.
Barry Sandrew, PhD
(He invented the first auto-colorization process, and he contributed to the Alice in Wonderland conversion)
(He thanked Pierre de Lespinois (from the previous panel. Pierre was very critical of conversion.) for the kind words. (sarcasm)) 70% of the over $80M of Alice opening weekend was from 3D screens. Rather than go through his slides, he showed recent work to present a more accurate sense of the current state of conversion.
Christopher Bond
Christopher oversaw conversion of Clash of the Titans, and presented the chronology of the process. Prime Focus received the call in mid-January to review the film. They had the kick-off meeting with the Director, Editorial, and VFX supervisors, who were skeptical that it could be done in 8 weeks, in London on Jan. 19th. The delivery date for the final version of the conversion was fixed at March 19th. The details of what they were to be given when work started included:
– not a locked cut / expect editorial changes / scene additions
– 90-100 minutes run time, 1000-200 shots
– No graded material, begin work on raw scenes ASAP
– Less than 1/8 of the VFX shots were final at the start of conversion
Having to deal with working with an unlocked, ungraded cut meant redefining their work process:
– Focus efforts on the “most locked” reels
– Work with raw scans (so they were prepped when the graded material arrived) / 8 frame handles head and tail
– 16 frames – 1935 shots = 21.5 additional minutes
– Conceive and develop tools to reprocess graded material once delivered
– Automate as much of the pipeline as possible
– Everything went into a database, which was created for this process
The creative elements came into play by this process:
– Define “Keystone” shots throughout the film
– Bring “Keystone” shots to final ASAP (approximately 5-10 days) and show them to WB
– Apply notes / feedback from WB and propagate to shots within the same sequence
– The dailies went through many rapid-fire iterations. Dailies were being worked on in 3 rooms, 1 Dolby and 2 RealD
– Because of the quantity of dailies, they came up with a grading system – A (perfect), B (some artifacts), C (first-pass)
– Twice weekly formal client reviews, progressing from scattered shots, to sequences, to reels, to, near the end, the entire movie
– Daily reviews in the last few days covered 15-17 minutes of reviews each day
– Toward the very end they applied convergence and watched the shots in ‘cut’
Lessons learned
– Do not underestimate editorial and ‘conform’ needs and reviews. This is a massive amount of data.
– Working with an unlocked cut meant that lots of content ended up on the cutting room floor.
– Clients change their minds, even on fundamental issues like how much stereo
– We can convert a movie in 8-10 weeks. Just stay calm.
Q&A
(Brad) Why is there so much controversy about conversion? (Warren) Consumers will vote with their wallets. (Barry) Something this new will attract both criticism and praise, but it will make money if done right. (Chris Bond) Sound, Panavision, TV all met the same criticisms, which faded as they improved and gained market acceptance.
(Brad) Will costs come down? (Chris Yewdall) TVs can do the ‘good.’ Avatar shows the ‘best.’ We will fill in the middle over time, and costs will come down.
(Brad) Do you see a head-end broadcast converter box? (Warren) Live conversion will not be as good as human intervention in the near term. The automated process in the TV cuts out the content maker from the revenue and control, so it is critical that they get involved in the process somehow.
(Brad) Chris (Yewdall), you brought up good, better, best. Will consumers see the difference? Should they be branded differently? (Chris Yewdall) No, they should not be branded differently. There are already variations; 720p v 1080p v 1080i. Over time, the conversion technologies will improve to the point where consumers won’t be able to tell the difference because there may not be a difference.
(Brad) What are the technical advantages of converting a new feature versus an old feature. (Barry) The degree of creative input is much less for older features. There is a great deal of creative input for new films, which adds time an cost. (Warren) We’ll have a new generation of filmmakers shooting in 2D, knowing it will be converted. Their shooting process will be changed by that knowledge, and the viewer will benefit from that. It is just starting to happen.