En
Fr
The birth of the project
Élie Levé : Momoko went to the École des Beaux-Arts and Le Fresnoy, a contemporary art school. She made her first student film, Planet A, in which she began using technologies such as time-lapse and mixing real and model worlds. Her third film in the Planet series, Planet Σ, was produced by Les Films d’Arlequin and Arte. This time she added high speed, and that’s when we met. I was working at a company called SubLab, which specializes in high speed and hybrid technologies. The film won the Audi Short Film Award in Berlin, was shown at hundreds of festivals, and received a dozen awards around the world. At that point, we got a request from Arte, which was starting to produce VR content. We adapted Momoko’s world to 360° on Planet ∞ , mixing high-speed cameras, time-lapse, and 360°. It was quite a challenge because we were working underwater with tadpoles, and at the time there were no 360° cameras, so we printed 3D rigs on which we mounted GoPros because they were the only tools that could go underwater. It was really a matter of improvising ! There were eight months of post-production. Just rendering a single shot took a week and three days of data transfer ! And then she started writing Planètes, ten years ago.
There was a lot of work involved in writing and financing, and on my side, developing tools. This time, we were bringing everything together in a single film, so there were a lot of different technologies involved : time-lapse, live action, animation, high speed, stack-shot (which is a technology that allows to artificially increase the depth of field in macro photography by moving the camera to take other photos and then combining them to accumulate all their areas of sharpness). We used this technology for the whole slug section. We scanned pine bark, which is about ten centimeters thick, to create 3D objects that could be assembled in post-production to create sets. I also worked with robotic arms coming from car assembly fabriks, whose internal digital components we redesigned so that they could be integrated into 3D environments. We developed a small version of these arms so that they could move, for example, 60cm, but over a period of three months in time-lapse.
We spent a year and a half preparing with Momoko. She was working on the script at the same time, so when she wasn’t with me, I would go directly to the developers, check out the equipment, and do tests on my own. We met up regularly to do photo tests, discuss issues with the film, camera angles, focal length, scale, and try things out in Photoshop to get an idea of the perspectives. It was quite a lot of preparation. We had to test every camera and every lens, and sometimes the equipment was only released as filming progressed, such as certain lenses and projectors. Sometimes we had to postpone certain scenes shooting because the equipment wasn’t released yet ! Some scenes were cut because they couldn’t be done live. There’s a scene I called the Valkyrie scene, with about fifty mosquitoes flying in formation to attack a slug. Filming one mosquito is complicated enough, let alone fifty ! So I told Momoko that this scene couldn’t be done any other way than in 3D, and when she heard that, she tended to understand that it was a scene that had to be cut.
Since the film combines so many different technologies, it is considered as an animated film, so we worked with Miyu Productions, a production company specialized in animation, who guided us through the specific processes involved in animation. In particular, they had us make an animatic, so the entire film was created in 3D before shooting, in order to have the framing, shot lengths, etc. figured out in advance. And it turned out to be necessary, in order to be able to assemble all our images, to know that the slug’s head movement should last a second and a half and not three seconds, etc. This animatic was made by Momoko and the editor, and I tried to stick as closely as possible to it on set. I showed Momoko what we could do technically, with the limitations of macro photography, which I tried to reduce as much as possible, but which were always there, in terms of depth of field, minimum focus, etc.
The shoot
EL : Filming took 260 days, spread over two and a half years, with periods of preparation and location scouting in between. The first part was in Iceland, from October to November. The idea was to film natural settings that we would then integrate into the backgrounds of our models, sometimes in motion. We couldn’t use tracking shots because we were on glaciers, sometimes with a 35-minute drive to reach the set, so the idea was to use drones. We started working with a drone operator on site, but on the first day of camera tests, we realized that the drone I wanted, the AltaX, which can support cinema cameras, was so big that it was much more susceptible to wind. So we went back on this idea and used smaller drones for certain scenes, particularly the cave interiors.
The second part was the longest, from January to August 2022, at the Château de Rambuteau in Burgundy. This castle has large 200 m² orangery rooms, where we set up seven time-lapse studios, in which all kinds of plants were growing, some against a blue background, others directly integrated into models we had built. When we arrived, nothing was suitable for filming, so we had electrical terminals installed, laid hundreds of meters of cable, brought in fiber optics, and created studios and systems for suspending cameras and lights. On the first day, the Count of Rambuteau told us, “If you want wood, there’s a tree that just fell”, so we went out with a tractor, brought the tree to the sawmill, and it became our film studio. We had about 20 cameras in this orangery, up to five per studio. It was a technical challenge because, obviously, the cameras couldn’t see each other, the lighting had to be uniform from all angles, and the plants, growing at a rate of several inches a day, couldn’t obstruct the cameras’ field of vision, so we had to check the plants’ progress almost every day. The plants also had to be watered without pouring directly water on it, otherwise we would have seen drops appearing on the image during the time-lapse, so we installed hydroponic systems in the models, which fed the plants from below. When we started shooting in January, it was almost -10°C in Burgundy, then the temperatures gradually rose to 35°C, so this seasonality dictated our schedule. We started with exotic plants, which we could shoot at any time of year since they were in a greenhouse, then we followed the rhythm of spring, the flowering of plants, the birth of certain insects, etc. We couldn’t afford to redo scenes since a plant only grows once !
We had a system that managed the time-lapse, with plasma screen filters that cut out the daylight for a second while the photo was being taken, and the rest of the time it restored the natural light that allowed the plants to grow. And all of this was synchronized with the flashes, the cameras, and the water supply to the plants. Each studio had its own box that managed all these triggers, which we had developed in preparation.
We had more than 30 models, weighing up to two tons. They started out by being crafted with wood, then were covered with plants by florists and decorators. Next, they went through the time-lapse phase, where the plants grew, and once that was done, we brought the models back to the studio where we shot the scenes with the insects. It included a conventional camera and a Phantom camera shooting at 1,000 frames per second, which required a lot of light, so the challenge was to find solutions to keep the temperature from getting too high for the insects. We put them in refrigerated tubes, which also allowed us to guide them a little as they moved. We also used a lot of projectors reflecting off mirrors to reduce the heat. It was quite complicated in terms of lighting, because we had to synchronize the light between the flash-lit cameras, the high-speed camera, which needed ten times more light, and the other cameras, which were sometimes shooting in natural light outdoors. The lighting conditions had to be synchronized for each shot. Fortunately, I could see the previous shots to make it easier to match. I could also remove the blue background live and see the sets we had filmed appear directly. That wouldn’t have been possible a few years ago ; we arrived at the same time as the technological tools that allowed us to make the film.
During preparation, we realized that when we filmed a slug wandering around and sped up its movement, it looked like it was sliding on the surface, which wasn’t realistic at all, so we decided to texture the floors we were going to shoot on to give a grippy effect. So, for some scenes, we used photogrammetry to extract the parts of the Iceland footage where the slug was supposed to walk, then we printed them in 3D and on a green screen at the scale of the slug. That way, by removing the green screen, we could paste the same slug back onto the rock where we had filmed six months earlier.
Seasonality was also important, especially for the bees : we only filmed bees that were going to die naturally within a day or two, so as not to disturb the colonies. We filmed with the slugs throughout July and August, and by the end we couldn’t find any slugs anywhere on the 200 hectares of the castle ! We were disappointed to have no more actors, and on the last day, while picnicking, we found about sixty slugs in the cool shade of a board, which saved the end of the shoot ! During this shoot, we spent hours and hours waiting for movement. We tried everything : music, mist, temperature changes, startling them, shining a light on them to try to get them to react. We made salad magic potions to make them follow a line from point A to point B. In two months, we had time to try everything and in the end we didn’t find a convincing solution ! We found the solution for other insects, creating tube systems with different temperatures so that they would move towards the coolest spot. It worked for butterflies and bees. Sometimes we were asked if we had called in specialists to work with the insects, but despite all their knowledge and years of research, they had never tried to put an insect in a refrigerated tube to go to the lights and pass in front of the camera ! No one had ever done it before, so it was only when the insect arrived in front of the camera that we could tell how it would react, whether the focus was right, whether the stimulus was right, whether we needed to use a fan to push it away, and so on. We just had to make the best technical preparations possible so that we were free to deal with the unpredictability of living creatures on set.
At the same time, we had other special effects studios to create smoke, flames, etc. That’s where the images of butterfly wings on fire were made : large models of wings were built and then set on fire, which we moved in slow motion with a high-speed camera. Then we superimposed the movement of real butterflies filmed in macro on the movement of the flapping wings.
We also had studios for the mushrooms. It’s quite complicated to work with mushrooms because you need to regulate not only the temperature but also the humidity, so we created biospheres that we could control ourselves. We even filmed phosphorescent mushrooms. We took a two-minute exposure photo to see the mushroom, and then we waited three weeks for the mushroom to grow so we could see what we couldn’t see with the naked eye. Even I couldn’t frame or focus with the naked eye ! I had to take a two-minute photo every time and wait to see the result. It was in caves, too, with bats attacking me in the dark. It was pretty epic !
We then shot in the Vendée region for the potato scene, with a huge boat. The process was to film ourselves with drones to get a real interaction with the sea, the feeling of the waves interacting with something floating on the surface. On the boat, the crew got seasick because we turned off the engine to really get the feeling of a dead body drifting in the waves, and it was the worst feeling, especially for the crew who were behind a screen without seeing the swell. The only solution I could offer was to feed everyone well, but I couldn’t do anything else !
Then there was the whole underwater part in Roscoff, at the CNRS. In Iceland, we scanned the sets with a drone, especially for the scene with the tadpoles in the waterfall, and then at the CNRS, we filmed underwater, placing all the anemones so that they would stick to the set filmed in Iceland. If we had wanted to shoot to scale, the anemones would have had to be the size of a truck ! So beforehand, we did the calculations between the drone’s focal length and the macro’s one and found the correlation to be able to place the anemones, with a year and a half between the two shoots. The problem with the submarine was that even when we did tests, we couldn’t simulate it in 3D technically because of the water effects. So we planned the robot’s trajectory, then we managed the scale of that trajectory as a percentage on the day of filming, adjusting the focal length and repositioning all the anemone, one by one.
So the robot arm was above an aquarium, and we needed a periscopic lens that would drop into the aquarium and then come back up at a 90° angle. All of the robot’s movements were reversed because the camera was upside down. The lens had just come out, and we realized when we used it that it was only waterproof for five centimeters, instead of the thirty centimeters that were advertised ! So we 3D-printed some diving suits for the lens, so at night the 3D printers were running, in the morning we put them on the robot and we managed to get waterproof optics like that.
We had planned underwater sets on which we could attach our anemones, which had to be heavy enough to sink to the bottom and not float, so we had created underwater grid systems, normally made of stainless steel, and on the first day, when we put them in the water, I saw the metal melting before my very eyes ! Even with the best preparation in the world, there are some things you just can’t anticipate ! We had to recreate everything on site, and for two days I went around hardware stores to create a new system.
We finished filming in Japan, where we shot time-lapse footage of moss and insects in the middle of nature. We exhausted all the local naturalists to find our stars ! I remember a scene with a large moth, which was magnificent. We had 25 scenes to shoot, we did the first 24 in two or three days without too many problems, and on the last scene, the moth escaped ! Since it’s a moth, it doesn’t fly very long under daylight and landed on a tree, ten meters high. So I had to climb to the top of the tree to get it and get our last shot ! We also wanted to film a dung beetle for the opening scene of the film, so we went to farms to collect cow dung and placed it at different altitudes because dung beetles live between 1,000 and 1,800 meters above sea level. We looked like crazy people !
Post-production
EL : Since we can’t precisely control the movement of a drone (because of the wind), the initial idea was to try to shoot the scenery beforehand, track it, then model it in 3D on the mock-up, and finally assemble the images in post-production using photogrammetry.
The only things in 3D animation are the dandelion seeds. There’s also the centipede in the storm at the end, which was done in 3D for budget reasons. But it was really Momoko’s original idea that there should be no 3D. We even did tests with dandelion seeds that were about a centimeter long, to try to make them move by blowing on them, but it wasn’t possible. With 600 shots, multiplied by four characters, while creating attachment and special characteristics for each of them, we couldn’t have done it. And it would have caused scale problems too. For example, we filmed the slugs on a green screen and then enlarged them to make them look a bit like dinosaurs, Jurassic Park style, so if we had wanted to have interaction between the dandelion seed and the slug, we wouldn’t have had the same scale ratio at all.
It also allowed us to get away with the depth of field you get with high-speed macro photography, where we would never have been able to get the bee in the distance and the plants in the foreground both clearly visible if we had wanted to do everything in one shot.
We had a Phantom camera, which was the central part of our images and which we really like with Momoko, and which I’ve been working with for a long time. Then we had cameras, which didn’t have the same depth of field with their large format sensors as the Phantom’s Super 35 sensor. We also had images taken by drones, and finally we had a Blackmagic camera, which was quite interesting because it’s really compact, which allowed us to suspend it from above and stick it to the set, and it filmed in RAW, and at the time, it was the only compact camera that could do that. So there were three different sensors, so beforehand we scanned all the sensors with Frank Malmin and Reepost, with the same lens, the same lighting conditions, with color charts and different sensitivities, and everything was archived, so we had a baseline reference for all the sensors and their digital noise information. And then there was the question of lenses. With 17 cameras, we couldn’t afford to have cinema lenses for nine months, so I decided to work with photo lenses for all the cameras, which allowed us to harmonize as much as possible, and the 45 different lenses were also scanned. So the most important thing was to announce the lens information on the clapperboard or by voice for each shot, so that post-production could retrieve the information and harmonize the images. This is because a single shot can contain up to seven merged images : the distant shot filmed with a drone, the model filmed with the Blackmagic camera, the time-lapse shot with the cameras, and the insect filmed with the Phantom camera. Each shot was pre-graded to match the four cameras, and then the shot was graded as a whole. The final color grading took three months, but in fact the entire process took two and a half years, as each craftsman in the post-production chain did a little bit of grading. I didn’t participate in the final color grading because I trusted Momoko and her photographer’s eye, which took over with her intentions. Technically, I knew that my work had been done upstream, to provide the best post-production chain I had anticipated in order to achieve the best possible color grading quality.
We ended up with about 1 petabyte of data, or 1,000 TB, in total with the copies, over three years of shooting. To make things a little more permanent, we decided to keep a copy of the film on LTO tapes. Another copy was stored at Reepost, on servers where they had a pipeline that everyone could access : the animators in Belgium who created the 3D dandelion seeds, the colorists, the sound engineers, etc.
I’ve always done this kind of project with technical challenges, in advertising or elsewhere, and I don’t really see myself in more conventional projects. I often help friends who are cinematographers with technical issues too. I was invited by the Chefs Op’ en Lumière festival, and I thought it was great to be able to talk to other cinematographers about the technical side of things that might scare them, so they could see that we have solutions to their problems and that it’s not as complicated as they might think.
(Interview by Margot Cavret for the AFC)
