For those who don’t know the tool: Nuke from Foundry is a node based compositing application used in feature film and episodic VFX. It ingests multi pass renders, deep data and USD scenes, and sits downstream of layout, animation and lighting while feeding DI and delivery.

Gaussian Splats go native

Foundry has released Nuke 17.0 with native support for Gaussian Splats. Artists can now import, view, manipulate, render and export splat data directly inside Nuke without third party conversion tools. Gaussian Splats are point based scene representations in which radiance is encoded as collections of anisotropic Gaussians rather than polygonal meshes. In Nuke 17.0, splats can be imported via GeoImport or GeoReference nodes and visualised in the Hydra based 3D viewer.

Rendering is handled through the SplatRender node, which supports depth output, deep output and motion blur. GeoDeletePoints enables non-destructive point removal, while GeoGrade allows colour grading operations on splat data. Field Nodes integrate with splats, enabling combination and transformation of point data. According to Foundry’s documentation, these nodes allow non-destructive masking and manipulation of 3D data, including Gaussian Splats.

USD 3D system leaves beta

The new USD-based 3D system is now fully released. It is built on Pixar’s Universal Scene Description and is intended to bring structured scene management into Nuke’s compositing context. A new import dialogue enables selective loading of data from complex USD scenes, including payload control and item-activation toggles. GeoImport supports USD with scene graph metadata as well as Alembic cameras and geometry. Axis nodes now include new constraint input pipes such as LookAt, Translation and Rotation constraints, alongside a Snapshot function to freeze transforms.

Camera workflows have been extended. Viewer locator visibility is improved, GeoCamera nodes can be created directly, and GeoEditCamera allows direct editing of USD camera data. Projection tools have moved into ScanlineRender. Cylindrical projection and UV Unwrap are supported, including UDIM tile outputs.

Material workflows have been aligned with USD and MaterialX. MtlXStandardSurface supports MaterialX Standard Surface preview and render in the Hydra viewer. New shader nodes include ReflectiveSurface for reflections and refractions, BasicSurface and WireframeShader. Legacy equivalents such as ConstantShader, FillShader and MergeLayerShader are retained.

Material binding is handled through GeoBindMaterial, which includes Binding Strength and Purpose controls to balance performance and render fidelity. Lighting nodes have been expanded. USD light attributes can be pulled directly into Nuke without baking. Updated DirectLight, SpotLight, PointLight and EnvironmentLight nodes include shadow toggles and output channel controls.

New USD aligned light nodes include GeoEditLight, GeoDistantLight, GeoDiskLight, GeoSphereLight and GeoDomeLight. Controls extend to intensity, exposure, falloff type, focus and colour temperature, with terminology aligned to physically based workflows. ScanlineRender now exposes explicit ray tracing controls, including Max Ray Depth and separate depth limits for reflection, refraction and diffuse rays. Motion blur controls include Shutter, Shutter Offset, Shutter Segments and Shutter Bias.

AOV output control has been expanded with Z depth, motion vector variants and independent surface AOV toggles. GeoMask and GeoClearMask nodes enable 3D masks to cascade down the node graph with visible indicators, improving readability in complex setups. GeoPython provides direct Python access to USD prims and schemas inside the scene graph, allowing scripted edits to USD data within the comp context. Graph Scope Variables can now operate inside USD scenes to swap geometry, cameras and textures based on variable logic.

BigCat scales machine learning

Nuke 17.0 introduces BigCat, extending the existing CopyCat machine learning toolset. While CopyCat focuses on shot level training, BigCat is designed for large scale dataset training across tens or hundreds of shots. BigCat supports data augmentation to generate variation for lighting and motion conditions. Validation tools include validation loss graphs and a validation contact sheet. Custom loss functions such as LPIPS are supported to encourage structural and semantic alignment beyond simple pixel difference metrics.

Graph Scope Variables mature

Graph Scope Variables receive a production ready framework update. Python callbacks and API integration allow custom code to be triggered on variable events. Root node knob expressions such as frame range and fps can drive dynamic variable values. Variable names can now be displayed in node labels for improved script readability. Foundry reports performance improvements for large scripts using Graph Scope Variables, though no numerical metrics are published.

Annotation and UI refinements

The annotation system has been overhauled. Drawing tools are more responsive, brushes are redesigned and a dedicated comment panel is introduced. Visual indicators aim to clarify review notes within the node graph.

In the 3D workspace, node colour coding, mask icons, scene graph filtering and live read indicators have been updated to improve clarity. Error and warning visibility in the viewer and node properties has also been enhanced. By the way, if that person looks familiar: You’ve read a whole story about that project in Digital Production:

We Hunt Giants

Core performance and colour

Core image processing updates include significant acceleration in TVIScale. Foundry states that GPU upscaling can be up to 98 times faster and CPU upscaling up to 26 times faster than previous behaviour. Deep composite rendering is reported to be up to 1.88 times faster to viewer and disk. Test conditions are not detailed in the documentation.

OFX plugins are no longer artificially resolution-limited. Plugins, including Furnace OFX, can now run at full hardware resolution. HDR MOV reading and writing now preserves YCbCr matrix, colour primaries and transfer functions. Native ACES 2.0 configurations are included. NotchLC codec support for MOV ingest and export is available on Windows and Linux.

Platform alignment and SDK updates

Nuke 17.0 supports USD 25.08 and aligns with VFX Reference Platform 2025. Third party libraries updated under this platform include Boost, OpenColorIO, OpenEXR, TBB and OpenVDB. OpenAssetIO and OpenAssetIO MediaCreation libraries are upgraded to version 1.0.0, with MediaCreation at 1.0.0 alpha.12.

Camera format support is updated via the latest Sony SDK and SMDK, including support for Burano firmware version 2 and 3.8K and 4K formats. Monitor out updates include AJA NTV2 17.5x, DeckLink 14.4 and NDI 6.x SDKs, with 10-bit support.

Parallel release for conservative pipelines

Alongside Nuke 17.0, Foundry has released Nuke 16.1. It includes most of the same feature updates but is built on the VFX Reference Platform 2024, allowing studios additional time to migrate their infrastructure.

Nuke 17.0 is available to customers with an active subscription. Pricing details are available from Foundry’s official product pages.

As with any major release that modifies 3D scene handling, machine learning workflows and core processing behaviour, facilities should validate performance, memory usage and pipeline compatibility before deploying to live productions. New tools and innovations should always be tested in controlled environments before use in production.

// Foundry Nuke 17.0 release notes
// https://learn.foundry.com/nuke/17.0v1/content/release_notes/nuke_17.0.html

And, if you want, you can compare it to the release of the Beta version in November:

Foundry Opens Nuke 17 Beta: New 3D System, New Variables

The post Nuke 17.0 rewires 3D and adds Gaussian Splats first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

Kraken is a Norwegian sci-fi thriller that moves between majestic fjord landscapes, industrial fish farms, studio interiors, LED-volume stages and CG-heavy underwater sequences. From a grading perspective, the main challenge was maintaining visual continuity while allowing each environment to retain its own identity.

The main priority for senior colourist Dylan Hopkin was establishing a strong, consistent foundational look across the film. “A good grade always serves the story and characters first,” he says. “What I do should be almost invisible. My colouring shouldn’t draw attention to itself.” Before the main grade began, Hopkin worked on teasers and the main trailer, a process that helped the team prepare for the actual movie, he explains. “But teasers and trailers are commercials for the film. The looks don’t always translate directly once you see scenes in their full context.”

Hopkin walks through a selection of scenes from the film, explaining how he approached colour, contrast and continuity, and how DaVinci Resolve was used to solve specific creative and technical challenges.

A bit about Kraken

Kraken (2026) is a Norwegian monster action feature directed by Pål Øie and produced by Nordisk Film Production AS, with Handmade Films in Norwegian Woods also credited on IMDb. It was shot on location around the Sognefjord and included studio work in Finland.

The story follows marine biologist Johanne, sent to investigate a salmon farm near Vangsnes after two teenagers die and strange phenomena suggest something huge is moving in the fjord depths.

Establishing a foundational look for Kraken

The main weight of the grade was creating a robust foundational look, which effectively became a bespoke show LUT implemented as a fixed node tree. From there, I focused on what each image already had to offer. I looked at dominant hues, how practical light flowed through the frame, what the most important part of the image was, and what the director and DP wanted the audience to feel. From that analysis, I usually get a gut feeling for how to adapt the image.

I try to be as respectful as possible towards the cinematography. What’s captured in front of the lens creates the DNA of the look. Creating appropriate tonal contrast so images sit right is a big part of grading. I do that using custom curves, DCTLs or different DRTs in Resolve, depending on the material. Shaping light is also critical. I rely heavily on windows, tracking and Magic Mask [more on which later] to guide the audience’s eye, in much the same way as dodge-and-burn techniques in still photography.

Maintaining continuity across scenes

This is one of the most challenging aspects of grading. I prefer to choose as few reference images as possible per scene and match everything else against them.

It’s also important to commit to a consistent set of tools. I often use colour-space-aware tools for exposure and white balance, such as Resolve’s HDR Grade tool, because the maths is very clean and respects the source material. Another approach I use is working in linear gamma with gain for exposure and white balance. The key is consistency.

Once shot-by-shot balancing is done, I use the Group Post-Clip node tree for scene-wide corrections. On Kraken, colour management was handled manually in a DaVinci YRGB project using nodes and input DCTLs. For animated features, I usually prefer ACEScct, but for live action, I still like a more hands-on approach.

Fjord landscapes

For the fjord exteriors, the goal was to keep foliage and skies natural but present. I relied heavily on the Tetra2 DCTL by Juanjo L. Salazar to tweak foliage hues, with very clean results. I also used the ColorSlice tool to add subtle density to greens without flattening highlights. Cloud texture in the skies was preserved through contrast control rather than saturation.

Kraken is the first Scandinavian feature film to be graded in Dolby Vision for theatrical release! For this additional version, Hopkin graded two days on Resolve with an advanced panel in Dolby’s reference theatre in London’s Soho. “It was an amazing experience: next level cinema,” says Hopkin.

Fish farm control room

The fish farm control room needed to feel grimy and industrial, in contrast to the high-tech control screens. It was shot in a studio with exterior views created using an LED volume, allowing precise control of time-of-day lighting while maintaining accurate reflections.

After establishing the base look, I added a subtle amount of green using primary offsets, similar to printer lights. That small adjustment added grit and helped sell the environment.

Fish farm exteriors: day and night

The fish farm exteriors had a weathered, oily quality with plenty of patina. Daylight scenes were graded with softer contrast and a colder Nordic feel.

At night, sodium vapour lighting dominated. I pushed the blacks deeper to add more bite and reinforce the industrial mood.

Kraken Underwater sequences and transitions

Underwater shots varied significantly, combining live action, CG, day-for-night and fully stylised sequences. Most daylight underwater scenes leaned green-cyan, while nighttime scenes moved more towards blue.

There were exceptions. During the nighttime kraken attack, surface lighting mixed strong sodium vapour with cooler sources. When submerged and looking up, I pushed highlights towards orange to connect the underwater shots to the surface action.

For shots moving between above and below the surface, I used keyframed grades in dynamic nodes to animate colour transitions. The HDR Grade tool, offset controls and hue-vs-hue curves were central here.

Submerged luxury restaurant

The submerged luxury restaurant was shot in a studio using an LED volume. I compressed the colour palette slightly using hue-vs-hue, managed saturated skin tones with hue-vs-sat, and controlled saturation depth with ColorSlice. The aim was restraint, keeping the environment immersive without overpowering performance.

In the restaurant, some wide shots suffered from LED moiré. I analysed which colour channels were affected and used a splitter-combiner node structure with Color Compressor, Debanding OFX and subtle directional blur, applied via qualifiers and masks.

Fish farm attack and fog

During the early evening fish farm attack, the Scandinavian light was at its best. I added subtle warmth along horizon lines and introduced fog behind Johanne using Cinematic Haze OFX.

I used Resolve’s masking tools and planar tracker to lock haze masks in place, excluding mountain silhouettes with HSL qualifiers. In total, I added fog and haze to around 40 shots.

In one canoe sequence, several shots lacked fog entirely. I tested Cinematic Haze, showed the results to the director, DP and producers, and ended up completing the remaining shots.

I like the control Cinematic Haze offers: finding the right depth-map sweet spot first, then adjusting airlight, density and resolution loss.

Laboratory bioluminescence

In the lab scenes, a green bioluminescent liquid needed to feel more prominent. I qualified the bioluminescence, added Add Flicker OFX for a smooth pulse, and used ColorSlice to push subtractive saturation into the green hue.

Red alert and the nighttime finale of Kraken

The red-alert sequence marked a deliberate escalation. This was the first nighttime attack on the fish farm, and the images needed to feel crisper and more aggressive than earlier scenes. Saturation increased due to intense red warning lights, but it was important to keep those colours under control. I relied on the Sat-vs-Sat curve to keep the reds vivid without tipping into a digital look, with gamut-mapping used where appropriate.

In the nighttime finale, Johanne leaves the fish farm by boat before the sequence cuts to day-for-night material. From that point on, the scene had to be transformed into a believable nighttime environment with a softer Scandinavian summer-night feel. I used Resolve’s HDR Grade tool for most of the heavy lifting and introduced a warm sodium-vapour tone to maintain a visual link to the fish farm. Tracked masks and Magic Mask helped preserve detail in Johanne’s performance while keeping the environment dark and moody.

Magic Mask and selective grading

Magic Mask has changed the way I work. Tasks like face relighting used to rely on tracked windows or rotoscoping, which could easily introduce halos. Magic Mask can select faces, clothing, vehicles and more with minimal setup. It’s not perfect every time, but with some tweaking, it solves complex selections efficiently. On Kraken, I used it extensively for relighting faces, isolating creatures and excluding actors from background colour manipulations, often in combination with keyers and windows.

Camera shake and lightweight VFX

In several action scenes, the DP Sjur Aarthun, wanted additional camera shake to increase impact. Solving this during the grade is far quicker than sending shots back to VFX. I used Camera Shake OFX for constant vibrations, and custom Fusion setups with damped-spring expressions for impact shakes, always with motion blur enabled. Working in Resolve allowed real-time review and better control over the scene’s flow.

Dylan Hopkin (IMDB | LinkedIn | Website) is a senior colourist based at one of Europe’s leading post-production facilities, Nordisk Film ShortCut Oslo, with more than a decade of experience specialising in high-end creative colour grading for feature films, TV dramas and commercials. He originally trained as a motion graphics designer before transitioning into finishing and online work.

Hopkin’s credits span award-winning cinema and television projects showcased internationally at festivals such as Sundance, Cannes Series, Tribeca and the Oscars circuit. He is also a certified DaVinci Resolve trainer, has created masterclasses on grading strategies and teaches at film schools across Europe.

The post Grading Kraken: insights from the colour grade first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

For those who don’t know the tool: DaVinci Resolve and Fusion Studio from Blackmagic Design cover edit, grade, comp and audio in one ecosystem. They run on macOS, Windows, Windows for Arm and Linux, with Desktop Video hardware for monitoring and dedicated control panels for tactile users.

Blackmagic Design has released DaVinci Resolve 20.3.2 and Fusion Studio 20.3.2. This is a maintenance update. No fireworks. Plenty of fixes. Sensible improvements in places that tend to bite.

The Resolve installer adds the main application, remote monitoring components and hardware control panel utilities. It also installs Blackmagic RAW Player and Blackmagic Proxy Generator. The free version of DaVinci Resolve 20 retains the same image processing pipeline as DaVinci Resolve Studio, supports unlimited resolution media, but limits mastering and output to Ultra HD or lower and supports only a single GPU on Windows and Linux.

Resolve Studio continues to offer multi GPU support, 4K output, motion blur effects, temporal and spatial noise reduction, multiple AI based tools, HDR tools, camera tracker, voice isolation, surround and immersive audio, additional Resolve FX, stereoscopic 3D tools and remote rendering.

Compatibility without time travel

Blackmagic states that project libraries remain compatible with DaVinci Resolve 19.1.4 at the database level. However, once a project is created or opened in 20.3.2, it cannot be reopened in 19.1.4. Facilities are advised to create full project library backups and individual project backups before upgrading. In shared environments, that advice is less a suggestion and more a survival strategy.

Edit page tweaks and trim discipline

A new preferences option enables dynamic trim in the Trim Editor on the Edit page. Dynamic trim adjusts edit points during playback, useful in dialogue heavy sequences and fast turnaroudn work. Subtitles receive more consistent kerning, a small but relevant change for broadcasters and OTT deliveries with strict QC.

Progressive renders are now supported for interlaced timelines with the same frame rate. Lens correction issues affecting vertical clips have been addressed. Media pool column header localisation issues are fixed, and slow media management startup in larger projects has been improved.

Magic Mask cache is now retained when pasting node attributes. Previously, pasting could invalidate cached analysis and trigger recalculation. The update reduces unnecessary GPU recompute in grading sessions that reuse node structures.

Faster AI scaling on Apple Silicon

SuperScale Enhanced is now up to 2.5 times faster when using the Apple Neural Engine. Blackmagic does not detail test conditions or hardware models. The claim applies specifically to SuperScale Enhanced and Apple Silicon systems with Neural Engine support. Not independently verified at press time.

Render speeds for side by side stereoscopic 3D renders have also been improved, although no numbers are provided. Timecode issues for proxies generated from stereoscopic 3D EXR files are addressed, along with incorrect render size estimates for some MXF clips.

Colour science and format fixes

Resolve 20.3.2 adds support for Fujifilm F Gamut C coloursapce and F Log2 C gamma. A ProRes RAW decode issue when using ACES has been fixed. Sony ARW decoding issues and incorrect pass through behaviour for some Sony XAVC renders are addressed.

Retention of MaxCLL metadata from XML when exporting IMF is corrected, relevant for HDR compliance. Several audio related issues are fixed, including problems with Dolby Atmos external renderers and an ADC reverb delay compensation issue in some projects. Support for Dolby headphone personalisation of binaural monitoring is added.

Fusion Studio joins the update

Blackmagic has also updated Fusion Studio to version 20.3.2. The installer adds Fusion Studio along with hardware control panel utilities, Blackmagic RAW Player and Blackmagic Proxy Generator. Fusion Studio 20.3.2 adds support for Fujifilm F Log2 C. It fixes issues with start and end render scripts on tools, resolves some character style problems in MultiText, and addresses an issue with the Import Footage menu. General performance and stability improvements are included.

Minimum system requirements across Resolve and Fusion Studio remain aligned with the current platform baselines, including macOS 14 Sonoma or later, Windows 10 Creators Update, Windows 11 for ARM on supported Qualcomm Snapdragon X Elite systems, and Rocky Linux 8.6 on Linux. Monitoring requires Desktop Video 12.9 or later.

Resolve 7 Resolve Studio 20.3.2 and Fusion Studio 20.3.2 do not introduce major new toolsets. They fix edge cases, improve consistency and add targeted performance gains. In other words, the sort of update most facilities actually need. As always, new tools and innovations should be tested before use in production.

// Quoted info
// https://www.blackmagicdesign.com/support

The post Resolve 20.3.2 tidies up the edges first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

Professional productions rarely consider smartphones. They lack ergonomics, reliability, and lens flexibility, and often process images in ways that destroy useful data before it ever reaches postproduction.We usually avoid “smartphone” news unless the device breaks into workflows in a meaningful way. But the legends at epic tutorials found something!

The Pixel 10 Pro may qualify. Unlike marketing claims around “computational photography,” this development deals with sensor-level access to true 12-bit video, a technical unlock that places this handset in a different category. The magic word here is Dual Conversion Gain (DCG), a hardware feature that has existed in many image sensors but has rarely been available for video capture outside cinema-grade gear.

What is Dual Conversion Gain?

Dual Conversion Gain (DCG), sometimes branded as Smart-ISO Pro by Samsung, allows a sensor to simultaneously capture high-gain and low-gain readouts of the same pixel exposure. The two signals are then merged before digitisation, producing a RAW output that preserves highlight headroom while also retaining clean shadow detail.

In practice, DCG produces one single file with:

• Higher dynamic range than standard single-gain readout.
• Finer tonal separation in shadow and midtone regions.
• Lower visible noise in dark exposures.
• Reduced artefacts compared to frame-stacked HDR video.

Conventional HDR modes, particularly in smartphones, often stack multiple exposures across frames. That introduces motion ghosting and temporal artefacts. DCG, by contrast, works at the pixel level and requires no stacking.

Interesting Detail: Google Pixel 10 Pro's Main Camera has 12-bit mode enabled by default!
byu/RaguSaucy96 inGooglePixel

Why is this important in the Pixel 10 Pro?

Previous Pixel devices, as well as Samsung Galaxy phones using similar sensors, already shipped with DCG capability. But manufacturers locked this functionality away. It was inaccessible to standard camera apps, and even advanced third-party software could not invoke it without root-level hacks.

The Pixel 10 Pro is the first mainstream consumer smartphone to expose DCG for RAW video capture directly to third-party apps. That means filmmakers, VFX artists, and technical directors can now record true 12-bit RAW video from a phone without voiding warranties or tampering with the operating system. Orjust ghavcing to deal with the shit-ton of work that is rooting a device these days. This change is not a marketing flourish, it is a concrete technical unlocking of a sensor feature that has been sitting dormant.

Testing and confirmation

Testing performed by developers using MotionCam Pro demonstrates visible improvements. Comparisons of waveform plots and histograms show increased usable dynamic range and visibly cleaner shadows when DCG is enabled. Noise levels in underexposed shots are drastically reduced. Blacks hold structure instead of collapsing into blocky artefacts. Gradients appear smoother, avoiding the typical banding that plagues 10-bit smartphone video.

Independent reports confirm that only the Pixel 10 Pro and Pixel 10 Pro XL models support this feature in video capture. The non-Pro Pixel 10 lacks the sensor bandwidth required.

Professional use cases

Let’s be clear: this is not a replacement for cinema cameras. But there are credible, professional-adjacent use cases.

BCAM/CCAM utility: For secondary angles, reference slates, or onset inserts where a lightweight, inexpensive camera suffices. The ability to record 12-bit RAW ensures that even imperfectly lit material has enough data for correction. But you have to consider how COlors and Contrasts land in, for example, Resolve – there is very little documentation, and you would need to test your footage to have a confident pipeline.

Photogrammetry capture: Tools like Meshroom rely on full-fidelity image data for reliable geometry reconstruction. Having decent video from a compact device could make mobile scans easier without elaborate lighting setups. Now, a LIDAR Scanner would be prefreable, but, in a pinch, with enough pictures, this should do as well.

Reference footage: VFX supervisors, colourists, or continuity staff may need to grab quick visual notes. Normally such reference clips suffer from baked-in compression and tone-mapping. A 12-bit file avoids this somewhat. And we are not talking about ProRes!

This makes the Pixel 10 Pro more of a pipeline tool than a narrative camera. Its footage can fit into established dailies or VFX workflows without breaking colour pipelines.

Cost vs. capability

By professional standards, the Pixel 10 Pro is inexpensive. A dedicated RAW-capable BCAM costs multiples more. That affordability opens opportunities: every team member could carry a RAW-capable capture device without straining budgets.

That said, ergonomics remain a limitation. Phones lack ND filtration, proper I/O, monitoring options, and reliable thermal management for long takes. Professionals should see this feature as an additional tool in the box, not a cost-saving substitution for proper gear.

Industry implications

The larger story here may be about smartphone manufacturers finally unlocking pro-grade sensor features. For years, hardware has been capable of higher-fidelity capture, but consumer apps and OS restrictions kept that potential hidden.

By making DCG and 12-bit RAW accessible, Google signals a shift. The Pixel 10 Pro may not replace cinema cameras, but it demonstrates that latent features can be delivered to professionals if vendors choose to. If this sets a precedent, future handsets may become more reliable companions in production environments.

The bottom line

The Pixel 10 Pro’s 12-bit RAW DCG capture is an interesting tidbit for professionals. Not a final-frame tool, not a marketing gimmick, but a useful side camera with real production/photogrammetrry/documentation utility. Treat it accordingly: a supplement for reference, technical, and capture workflows—not a substitute for your Alexa, Venice or URSA. And we ignored the questions of thermal stability, variable Framerate, standardized colour profiles, and so many more things that make a professional camera and a smartphone two different beats entirely.

A reminder

Innovations like this should always be tested thoroughly before use in production. Smartphone footage may now hold up in a certain few workflows, but reliability and stability remain critical factors.

The post Pixel 10 Pro unlocks true 12-bit RAW with DCG first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

DP: Let’s kick things off with a bit of background: What actually set you and your crew on the path to AI relighting? 

Hoon Kim: Four other technologists and I founded Beeble as a selfie relighting app after seeing a gap in the market for AI-powered lighting tools. It received a great reception and went viral, but the real Eureka moment happened next. We found that VFX artists would hugely benefit from a similar tool that relights video instead of still imagery. We knew it would solve many time and budget issues, while enabling professionals to fix lighting in post, without having to reshoot. 

DP: With Switchlight 2.0 being released versus 1.0: What’s the “wow, that’s new!” factor? 

Hoon Kim: Switchlight 1.0 already had a huge level of detail in terms of PBR map generation, but this update brings it up a notch. It boosts the quality of our normal, basecolor, metallic, roughness, and specular passes, thanks to training on 13 times more data. We’ve now also extended these maps to background objects and environments, as well as the central video subject. 

DP: For anyone curious to jump in: How does the pricing and licensing for Switchlight 2.0 work?

Hoon Kim: All our pricing is credit-based, giving you a set amount of content to process each month. The Starter plan is free and allows you to run 15-second video passes, ideal for testing the software. For ongoing use, the Creator plan is $19/month and the Professional plan is $75/month, with discounted rates available on annual subscriptions. All pricing details are available on our site.

DP: If you think about Switchlight landing on the desk of a busy VFX artist or colorist: Where does it shine? 

Hoon Kim: Put simply, it delivers the level of quality a VFX artist would expect. Additionally, they receive AOV passes almost instantly, a significant shift from the traditional, manual CG rendering process, making it ideal for quick fixes.

DP: Deflickering temporal noise is a game-changer for a lot of us! Can you give us a peek into how your new approach works, especially for fast-paced or multi-source lighting?

Hoon Kim: SwitchLight 2.0 operates on an image-based model, utilising PBR maps that mitigate temporal noise deflickering. Our next update, SwitchLight 3.0, takes things further: a native video model that removes the need for special deflickering. More details coming soon!

DP: Speaking of standards: What’s your support like for things like ACES, Rec. 709, P3, and other pro color spaces? Are you outputting true 16-bit or 32-bit float EXRs, so no one’s losing data along the way?

Hoon Kim: Our AI is trained on the Rec. 709 colour space, which is the recommended format for best results. That said, you can still apply the generated PBR maps to your original ACES footage without losing any detail.

DP: What kind of shots or material have you seen Switchlight absolutely nail and where does it still sweat? 

Hoon Kim: SwitchLight 2.0 is built to handle footage from any camera, whatever the quality, lighting, or style. Of course, high-definition, steady shots with flat lighting will always give you the greatest control.

DP: When you’ve got those nightmare shots with crushed blacks, low-res and ancient TV rips: Do you have any hacks? 

Hoon Kim: For that kind of challenging footage, it helps to upscale before running it through Beeble, and then apply a separate deflicker pass on top of the generated PBR maps. Tools like Nuke Copycat make that step straightforward and can dramatically improve the result.

DP: Any “dos and don’ts” from what you’ve seen in the wild?

Hoon Kim: AI upscaling is definitely a “do”, it could massively improve poor-quality footage and makes it easier for SwitchLight to generate accurate PBR maps. But avoid heavy denoising, which strips away detail the AI needs, and hold off on strong LUTs or oversharpening until after relighting, as they can distort the data and lead to unstable results.

DP: Is there a “Switchlight for Dummies” checklist for prepping shots? 

Hoon Kim: We actually put together an article on this! However, my main advice is simple: aim for even, level lighting when shooting. Beeble can handle almost any condition, but flat lighting gives you the most flexibility and control later in post.

DP: Alright, time for a technical deep dive! How does Switchlight 2.0 whip up those depth maps? 

Hoon Kim: At its core, it’s a neural net doing the heavy lifting. That’s about as far as I can go! You can more information via these links: https://beeble.ai/research/switchlight-2-0-is-here or the paper on the technology: SwitchLight: Co-design of Physics-driven Architecture and Pre-training Framework for Human Portrait Relighting

DP: Under the hood, what kind of AI are we talking about? 

Hoon Kim: We are always improving the network architecture, so “monster hybrid” would probably be the most accurate description. However, the beauty lies in integrating the Cook-Torrance BRDF model, which enables the AI to learn how light bounces off a subject. Paired with physically based rendering, the system can generate a relit scene that looks physically convincing.

DP: The classic pain point: specularity. How does Switchlight tease out reflections and highlights? 

Hoon Kim: We’ve made specular maps flexible enough that you can dial them in however you like. Want more gloss, less sweat? Just tweak the strength directly in-platform, or plug the maps into a third-party shader like Blender or Unreal Engine for full control. You can even isolate high- or low-specularity regions, so you decide exactly how shiny or matte things should look.

DP: What about your reference data? 

Hoon Kim: All of our training data is created in-house using scans of real humans and objects. We train only on safe, clean, and fully licensed datasets, ensuring professionals can use our service with complete peace of mind. We do not use movies, films, or any third-party content that could raise copyright concerns..

DP: Platform time! So you’re solid on Windows, Red Hat, Ubuntu. Any dreams of a Mac version for the “Final Cut Victims”? 

Hoon Kim: Right now, SwitchLight 2.0 and the Beeble platform are web-based. With 3.0, we plan to introduce a local version, one that better aligns with the security and workflow standards of major post-production studios.

DP: For anyone building a new Switchlight workstation: What’s the minimum gear to get real-time, and what’s your dream setup? 

Hoon Kim: We’ve designed Beeble to be accessible to anyone, whether you’re a content creator on a laptop with phone footage or a studio running complex plates. At the entry level, all you really need is a laptop and some footage to upload; the platform handles the heavy lifting and generates the VFX passes for you. Bigger studios will naturally push higher-res material through, but it’s not a requirement. And for anyone wanting to get started right away, we’ve just released Beeble Camera: a free app that brings camera tracking to Beeble.

DP: Looking into the crystal ball: Where do you see Switchlight and AI relighting going, say, five or ten versions from now?

Hoon Kim: Looking ahead, our goal is to keep pushing the detail and fidelity of the PBR maps while advancing motion capture, camera tracking, and depth generation. Ultimately, we want to break footage down into its most detailed components so every frame can be reconstructed, relit, and reimagined with complete creative freedom.

DP: For the tech-heads and code nerds: Are there research papers or blog posts you’d absolutely recommend to get your head around the science powering Switchlight? 

Hoon Kim: Absolutely. Our paper is here for anyone to read. It provides a detailed overview of the methods behind SwitchLight, from self-supervised training to the physics-based models that underpin our approach. 

DP: What’s the most fun or jaw-dropping thing you’ve seen Switchlight do that made you stop and say: “Okay, that’s actually impossible—except now it isn’t”?

Hoon Kim: A standout for me was a father-and-son duo who made a fan film called Star Wars: Entrenched, with Beeble helping along the way. Navaz Dowling, who runs Bad Beetle Entertainment, taught his son the basics of VFX and brought in friends for costume design and relighting. The trailer’s already out, and the post-production community has responded really positively.

At the same time, we’re seeing larger post-production studios integrate Beeble into their pipelines. The work Boxel Studio did on Superman & Lois is a great example of this. With SwitchLight 3.0 introducing a local version, that side of adoption is only going to grow.

The post We have to talk about Switchlight 2.0 first appeared on DIGITAL PRODUCTION and was written by Bela Beier.