《Project Hail Mary》：Framestore FPS 如何把虚拟摄影机控制权交给 Greig Fraser

There is something rather appealing about giving a cinematographer a dial to move the sun. On Project Hail Mary, that was part of a much larger effort to make a digital environment respond like a film set. Greig Fraser could explore a spaceship, choose a lens, adjust exposure and find a composition months before the physical sets existed. The technology worked because it accommodated the way he wanted to shoot.

For this fxpodcast, we do a deep dive with Framestore Pre-production Services (FPS) Realtime Supervisor Michael Zaman and Virtual Production Supervisor Connor Ling about their collaboration with Fraser. Directed by Phil Lord and Christopher Miller, Project Hail Mary adapts Andy Weir’s novel, with Ryan Gosling playing science teacher Ryland Grace on an extraordinary mission to save Earth. FPS spent roughly a year at Shepperton Studios helping the filmmakers turn that adventure into practical decisions about sets, cameras and light.

The team’s involvement began within production designer Charlie Wood’s art department, supported by VFX. The ship’s interiors would be built as separate sets, so their relationship needed to be understood before anyone could physically walk between them. A virtual version allowed the filmmakers to explore Grace’s journey through the complete spacecraft, assess scale and consider how its changing gravity would affect their understanding of the space.

VR also let Fraser and the art department test focal lengths and camera positions, identify structural obstructions, and flag panels that needed to be removable before construction. Discovering that a wall prevents the intended shot is considerably more useful while it remains a design decision.

FPS made those sessions collaborative. Mirroring a headset directly to a monitor produces a restless, wide view that can be uncomfortable for everyone watching. Ling’s team attached a virtual cinematic camera to the headset, with controlled filmback and lens settings, then stabilised its position and rotation. The person in VR could explore freely while colleagues watched a readable view of the same decisions. An artist and operator supported the session, allowing changes to be discussed and made immediately.

That spirit carried into the motion-capture volume, where the filmmakers blocked action with performer Josh Wichard before exploring it through virtual cameras. As Ling puts it, the work was “always as information, never to tie them into a direction.”

Fraser’s involvement pushed the tools towards the language of cinematography. Lighting discussions used lumens, ISO and apertures. Virtual fixtures represented lights intended for the physical sets. The digital environment needed to give him a useful basis for judging exposure and direction, with enough control to shape the image creatively.

“Digital can do anything, that doesn’t mean it should,” observes Zaman.

FPS built a bespoke physical mixing deck with tactile dials and trackballs, and adapted its shoulder-mounted virtual camera to Fraser’s requirements. Exposure and the virtual sun became controls he could operate through familiar physical gestures. He could move through a scene using joysticks, adjust sunlight along individual axes and explore the relationship between ship, planet and illumination. For the space sequences, the team calculated astronomical relationships as a starting point, then gave him room to art-direct the result.

The controls helped sustain a familiar working rhythm. Focus could be pulled with a physical focus, iris and zoom controller linked to the engine, while another operator adjusted lighting. Multiple frame guides let Fraser consider different presentation formats as he composed. Across 65 days in the volume, he captured 1,096 virtual-camera shots. These were filmed choices about movement, framing and light, recorded with data that could travel onwards into VFX.

The film’s lens package added a distinctive complication. Zaman describes an ALEXA 65 paired with vertically oriented anamorphic lenses for the space material. That configuration contributed to the tall IMAX framing and unusual optical character, but required FPS to adapt its virtual filmback calculations for the direction of the de-squeeze. A conventional virtual camera setup could not simply stand in for the chosen optics.

Those choices became more demanding once the work moved onto the physical stages. FPS supplied content and operators for an Amazon MGM Studios Simulcam system, providing live composites of the photographed set and its digital surroundings. Fraser could see how a partial spacecraft fitted into the larger environment, adjust engine lighting and judge whether the practical illumination needed additional bounce or a change of direction. The same assets and controls developed in preparation remained useful during photography.

RedSpy supplied camera tracking through a sensor mounted on the camera and reflective markers installed around the stage. Tracking established where the camera was in the room; LiDAR supplied the relationship between that room and the physical set. By loading the point cloud into Unreal Engine and aligning it with tracking references, FPS could register the virtual extension to the photographed scenery.

Lens information completed the picture. Focus, aperture and zoom data, where applicable, were needed alongside camera position. Rehoused vintage anamorphics introduced further changes as focus moved, and the vertical orientation complicated calibration again. The live composite had to respond within milliseconds while the camera crew continued working. The tracking arrangement also needed to survive ordinary production changes, including a planned jib shot becoming handheld at short notice.

Maintaining the resulting data mattered as much as displaying the composite. FPS used FarSight Studio to record virtual-camera properties and preserve scene states for individual takes, including asset versions and lighting changes. With the sets evolving throughout preparation, a camera move alone would not have been enough to reconstruct an earlier decision. Custom tooling supported the hybrid Simulcam setup and its handoff to post.

The unusual optics made these records particularly valuable. Zaman recalls automatic tracking solves producing implausible values when confronted with the vertical anamorphic material. Recorded camera and lens data provided a trusted reference for resolving the shots. The system preserved the conditions under which an image had been made, giving downstream artists something concrete to work from.

Editorial was involved throughout. During the volume shoot, FPS delivered rushes twice daily, handing over the morning session at the break and the afternoon session after shooting. Simulcam composites also passed from video village to editorial. Zaman recalls approximately 2,000 video files and 60 hours of footage from the virtual shoot. Some early material stayed in the cut well into post-production, allowing timing and storytelling to develop while the final images were being built.

The work continued beyond Shepperton. Fraser returned for further virtual-camera sessions, and FPS also captured additional sequences alongside editorial in Los Angeles. Camera animation, lens settings and lighting information were supplied to VFX so that fully digital shots retained his decisions. As Zaman explains, “The goal was to render what he shot, not to change it at all.” Adjustments remained possible, but the starting point was a shot already photographed by its cinematographer.

FPS also carried the process into postvis, particularly in refining Rocky’s performance inside his travelling sphere. The practical puppet supplied mannerisms and lighting reference, while digital exploration helped resolve actions that were difficult to achieve physically. Editorial could assess the character’s intention and emotional clarity before committing to final rendering.

The practical lesson from Zaman and Ling is that successful virtual production depends on the working relationships around it. The tools developed with the production, supported by heads of department who understood their value and were willing to use them early. FPS gave those departments a shared space in which to make decisions, while retaining the flexibility to change course on set.

For Fraser, that meant extending his camera package into places no physical camera could reach. He could still choose the lens, find the frame and decide where the light should fall. Sometimes he simply had a rather longer reach to the sun.

Listen and/or watch our complete fxpodcast conversation on YouTube or wherever you download your fave podcasts.