For those who don’t know the tool: OctaneRender is a GPU renderer that sits after lookdev and before finals, and it talks fluently with DCC plugins plus OpenUSD and MaterialX, when it is in a good mood.

The alpha, the scope, the fine print

OctaneRender 2027.1 Alpha 1 is out as a public development build intended for testing by experienced users, and it is explicitly not positioned as production ready. Scenes saved in this alpha are not guaranteed to stay compatible with future builds. The build is Windows only right now, the minimum required NVIDIA driver version is 572. The downloads listed for this alpha are the standalone Windows installer and a ZIP archive. Other build types are planned for subsequent releases.

Real time neural viewport, now with thunderbolt energy

The headline feature is a new real-time neural viewport mode that replaces the previous real-time mode. It is built on the neural radiance cache framework introduced in OctaneRender 2026. The alpha describes the new mode as enabling near noise-free interactive rendering at very low sample counts compared to traditional path tracing. The mode is designed to preserve the spectral path-tracing behaviour while improving interactive feedback, even under complex lighting conditions. It combines internal research with NVIDIA DLSS Ray Reconstruction.

In the UI, the alpha notes a dedicated toggle. The real-time neural viewport can be enabled or disabled using a Thunderbolt button in the viewport components bar. Quality settings for the real-time denoiser live under the imager node, with the goal of balancing performance and quality depending on the hardware. If you have ever watched a veiwport go from soup to something readable right as a director walks over, you already know why this matters. The promise here is faster clarity, not just faster pixels.

Upscaling joins the party: DLSS and FSR modes

This alpha adds support for new up-sampling modes based on NVIDIA DLSS and AMD FSR. The build also references AMD FSR 3. The implementation sits in the Upsampler section of the imager node, alongside other up-sampling methods. On the DLSS side, the referenced technology page describes DLSS Ray Reconstruction as an AI method that replaces hand-tuned denoisers with an AI network that generates higher quality pixels between sampled rays in intensive ray-traced scenes. Those statements describe DLSS generally, not a specific Octane integration result. This renderer is CUDA based, and the Windows alpha focuses on NVIDIA GPUs (Big surprise!).

AI Light 2.0: sampling that tries harder

AI Light 2.0 lands as a new light sampling method intended to resolve noise faster. AI Light is described as a hierarchical algorithm for sampling scenes with many light sources, and the update is positioned as a major revision compared to the original introduction back in OctaneRender 4.0.

For lighting heavy setups, this is the kind of change that can alter how fast you converge on creative decisions. If your daily routine includes juggling dozens of practicals, emissive set dressing, and a few too many hero gobos, any sampling improvement tends to show up right where it hurts, in the first minutes of iteration.

USD export and a more hands on USD workflow

OpenUSD workflows get a concrete upgrade. Geometry export to USD is included, and the build adds the ability to export a scene in USD format. The cgchannel overview also notes that geometry export in USD can include limited support for USD Preview Surface materials.

On the editing side, the standalone edition includes a USD Stage editor that allows selecting variants and layers without saving and reopening the USD file. The alpha also adds built in support for authoring variants and muting USD layers directly within the USD geometry archive.

If your pipeline already treats USD as a transport layer between layout, lookdev, lighting, and downstream tools, these changes aim straight at friction. More export options reduce the number of awkward handoffs that turn into late night chat messages that start with “hey quick question….”

UVs and geometry: more room for messy reality

The UV ceiling changes, because production UVs are never as polite as tool demos. The maximum number of UV maps allowed per primitive increases to 10, up from 3, and UV sets can be assigned arbitrary names in DCC apps that support it.

Other geometry and attribute changes include support for named UV vertex attributes, support for using vertex attributes in bump maps, support for colour vertex attributes in vertex displacement, and a Vectron tube primitive addition.

If you have ever inherited an asset with five UV sets and a note that says do not touch UV4, you now have more breathing room. The named UV support also helps keep humans oriented when the same mesh carries multiple UV purposes across shading, decals, and baking.

Changelog nuggets for TDs who read release notes for fun

Beyond new features, the alpha lists a spread of resolved issues and smaller changes. A few examples include fixes around DDS preference crashes, Alembic instancing, several MaterialX node behaviors, and multiple OpenUSD import issues such as points not loading correctly and animated point instances importing incorrectly.

The Lua API gets deprecations in geometry exporter properties, including moving from renderSizeX and renderSizeY to textureRenderWidth and textureRenderHeight, and replacing writeOcsData with embedNativeData. A small UI level tweak removes Clock Rate and Cores fields from the GPU information panel. This is where you start thinking about tool glue. If you have scripts that poke exporter properties or automate export settings, the deprecations are your early warning sign to check for breakage.

Availability, licensing, and pricing

This alpha is available alongside the stable OctaneRender 2026 releases. The software is offered as rental only via Studio+ subscriptions. The pricing listed is €23.95 per month or €239.88 per year, and the Studio+ subscription page also lists an annual price presented as 19.99 € per month billed yearly.

Free Prime editions are also available for OctaneRender and Octane X, with the Prime editions limited to rendering on a single GPU and offering a smaller set of DCC integration plugins. The alpha download links themselves require a licensed customer login for access. That is where OTOY enters the picture in the most practical way possible.

Production reality check

This build is explicitly framed as experimental, and it is not recommended for production use. That guidance aligns with the normal alpha reality: features change, files drift, and stability work continues in later builds.

New tools and innovations should always be tested before use in production, ideally on representative scenes with your actual assets, shading conventions, and deadline pressure.

What to watch while you test

The big workflow test is whether the neural viewport mode makes iteration measurably smoother on your scenes, and whether AI Light 2.0 shifts your sampling strategy. The second test is whether USD export plus USD Stage editing reduces round-trip times in YOUR pipeline. The third is whether the DLSS and FSR upsampler options behave predictably inside your lookdev and review loop, especially when you rely on consistent framing and visual stability.

https://render.otoy.com/forum/viewtopic.php?t=85551

The post OctaneRender 2027.1 alpha goes neural first appeared on DIGITAL PRODUCTION and was written by Manuel Kotulla.

For those who don’t know the tool: Unity is a realtime game engine that ships to desktop and consoles, and Platform Toolkit sits in the build and platform integration layer, so your release pipeline spends less time arguing with targets and more time making frames.

What changed and when

A Unity product update at the Game Developers Conference included two platform-focused announcements. First, official Steam support is planned, including a native solution intended to help bring games to the platform. This support should be covering Steam, Steam Deck, and Steam Machine.

Second, targeted enhancements to the engine Linux runtime are planned with the goal of native performance increases. The same update frames this work as a way to reduce reliance on running Windows builds through Proton.

https://twitter.com/unitygames/status/2031777294394552663

Steam support without the scavenger hunt

If you have shipped on Steam before, you already know the punchline: it is possible, but the path has historically depended on what you integrate and how you maintain it over time. The new message is about making that experience feel like a supported lane rather than a community tradition. The announcement describes a built-in platform toolkit that will facilitate the workflow of bringing games to Steam. That phrasing matters for production because it points to changes that sit inside your day to day build and release loop, not just documentation or a sample project you copy once and then forget until it breaks on a Friday.

So for now, the important production takeaway is narrower and still useful: the roadmap points at a native solution for Steam targeting Steam Deck and Steam Machine as part of the same effort.

Linux runtime work aimed at less Proton reliance

The second half of the update calls out targeted enhancements to the Linux runtime in Unity, framed around native performance gains and removing the need to rely on Windows through Proton.

For teams shipping to Linux ecosystems, that focus is notable because runtime changes tend to land below your gameplay code and sometimes below your rendering decisions. When runtime behavior shifts, it can affect performance, stability, input, windowing, threading, and the kinds of issues that only appear on particular hardware and driver mixes.

Still, the direction is clear enough to influence planning. If your release strategy currently depends on Windows builds plus Proton, the announcement signals that native Linux performance is a priority area rather than an afterthought. That can change how you triage bugs, where you invest profiling time, and how you communicate platform expectations internally.

What is still missing

The story is big, but the specifics are currently thin in the provided material. There is no published schedule in the sources for when official Steam support will ship. There is no description of what the native solution includes, whether it is a package, an editor workflow, a build target configuration, or a set of services. There is no compatibility matrix that spells out editor versions, player versions, or which OS targets are covered.

That means any team looking to act immediately should treat this as a roadmap signal rather than a shipping feature drop. It is a green light to pay attention, not a license to rewrite your platform plan overnight.

https://unity.com

https://gdconf.com

The post Official Steam support lands on Unity roadmap first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

For those who don’t know the tool: GIMP is a free, open-source cross-platform image editor for painting and comp work, and it plays nicely with Inkscape when you need vectors and pixels to stop bickering. Think Blender for images.

Momentum, but with fewer faceplants

GIMP 3.2 ships as the first 3.2 release dated 2026 03 14, following a year of design, development, and testing by volunteers and the community, with an explicit plan to streamline releases after 3.0. The headline theme is non destructive layers. Text layers already lived that life. Now Link Layers and Vector Layers join the party, and they come with a new safety latch: you must Rasterize before destructive edits like painting. When you regret that choice, Revert Rasterize clears destructive edits and returns the layer to its non destructive state. That little switch quietly saves hours and reputations.

Link Layers let you pull an external image into a composition and scale, rotate, and transform it without losing quality or sharpness. The layer updates when the source file changes, and you can create one via Open as Link Layer in the File menu. You can also swap the linked file later in Layer Attributes.

Vector Layers arrive through the Path tool. You can draw shapes with adjustable fill and stroke settings, and the vector shape updates as the path changes. Non-destructive rotate, scale, and transform apply here too, so the shape stays editable longer. Yes, we like GIMP, and it is fun to watch the momenutm build when layers start acting like they understand revision notes.

Filters stack higher, and channels get invited

Non-destructive filters now apply to channels, not just layers and layer groups. They also apply to the new non-destructive layer types, so you can keep edits flexible without immediately baking them in. Layer groups get a neat trick: an empty group set to Pass through can carry non-destructive filters that affect all layers below it. Stack multiple filters on that group and you get an adjustment style workflow that stays tidy instead of turning into a layer salad.

A new Vibrance filter lands via updated GEGL. It combines Saturation and Hue Chroma behavior for selective saturation, with a specific goal of keeping skin colors from getting over saturated. You can reach it through the GEGL Operations tool or by searching actions with slash. The release also ships companion updates for babl and GEGL as part of the wider toolchain refresh.

New tools and innovations should be tested before use in prdouction, especially when they touch core layer behavior and color pipelines.

Brushes, text, and other daily annoyances get nerfed

A new paint mode called Overwrite replaces color values as you paint without blending alpha values together. It targets pixel art workflows, and it pairs best with the Pencil Tool.

The MyPaint Brush tool updates to the newer Version 2 API from MyPaint. That upgrade adds 20 new brushes, including an arrow brush, grouped under the Dieterle tag. Text editing gets less in-your-face. The on canvas Text Editor can be dragged out of the way while you type, and it supports common shortcuts like Ctrl plus B, Ctrl plus I, and Shift plus Ctrl plus V for paste unformatted text.

https://peer.tube/videos/embed/64DzkCtrEw9aDr88urpAnj

Exports get more grown up

Format support expands, including built-in support for more formats and export support for JPEG 2000, a standard for digital cinema. Export support is added for PSB large document files, matching existing import support. The DDS plug-in can export BC7. The OpenEXR plug-in can render Luminance Chroma images in full colour rather than grayscale. Vector Layers also affect export: SVG and PDF export can include raster and vector layer data together. Import support is added for AVCI (a HEIF variant with H.264 images encapsulated in HEIF).

The small print, literally and figuratively

On Windows, the downloads page lists supported OS as Windows 10 1903 or newer. Now Windows 7 anymore, and that makes you sad, take loooong look in the mirror. On Linux, distribution options include AppImage, Flatpak via Flathub, and Snap via the Snap Store.

GIMP 3.2.0 is the last version distributed as 32 bit on Windows. Starting with 3.2.2 and later micro versions, there will be no 32 bit Windows build. One side effect is that the 32 bit TWAIN plug in for scanning on Windows stops being distributed, with guidance to use manufacturer scanning software for now and an invitation to help build a 64 bit WIA plug in. Pricing is free, and open source – biut you can always donate, they REALLY have earned some help!

https://www.gimp.org/release-notes/gimp-3.2.html

https://www.gimp.org/downloads/

The post The GIMP Team boosts GIMP 3.2 first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

For those who don’t know the tool: Redshift from Maxon is a GPU renderer for Cinema 4D, Maya, 3ds Max, Houdini and Katana. It targets final-frame GPU rendering and USD Solaris pipelines, sitting at lookdev and lighting in production.

Consolidation, not features

Maxon has released Redshift 2026.3.0. The build includes plugins for Cinema 4D, Maya, 3ds Max, Houdini and Katana. It is a maintenance release focused on stability, host compatibility and incremental displacement updates. No new headline rendering features are introduced.

On Apple hardware, Redshift now requires macOS Sonoma 14 or higher. Maxon recommends macOS Tahoe 26.2 as the latest officially supported version, while noting known rendering artefacts on Intel CPU plus AMD GPU systems on Tahoe. On Windows and Linux, updated NVidia driver minimums apply, including specific requirements for Blackwell 50xx GPUs. Linux requires glibc 2.28 or higher.

Texture displacement cleaned up

Texture displacement receives performance improvements and several correctness fixes. The update resolves holes in meshes, vertical artefacts in the software ray tracer and cases where displacement disappeared on tessellated meshes without UVs.

Support for smooth normals has been added to texture displacement. The UV Context Projection node gains a new camera selection mode for camera projection mapping.

Additional fixes address deformation motion blur with displacement, incorrect auto-bump behaviour, and noisy details on scaled meshes. Material blending issues involving stacking and the Material Blender shader have also been corrected.

Hydra and Solaris updates

Redshift adds support for Houdini 21.0.559 and 21.0.596 and drops support for 20.5.654 and 21.0.440. The Hydra delegate now supports the Solaris 21.0 statistics panel in the latest Houdini 21.0 production build.

Hydra 2.0 fixes include motion blur limitations, primvars sampling issues, light linking problems, nested instancing errors and live updating of custom light and camera parameters. Support has been added for USD implicit surface primitives, USD mesh lights via the Geometry Light LOP, and updated USD and Alembic procedurals. AOV-related bugs affecting ObjectID, Depth, cryptomatte, deep and Refractions Light Groups have been fixed.

Redshift 2026.3 is a stabilisation release aimed at removing pipeline friction introduced in earlier 2026 builds. As always, new tools and innovations should be tested before use in production.

// Redshift 2026.3.0 Release Notes
// https://support.maxon.net/hc/en-us/articles/Redshift-2026-3-0-2026-02-February-10-2026

The post Redshift 2026.3 fixes displacement headaches first appeared on DIGITAL PRODUCTION and was written by Bela Beier.

For those who don’t know the tool: DJV is an open source image sequence viewer and playback tool for VFX and animation. It runs on Linux, macOS and Windows, supports OCIO colour management, and is typically used for dailies, shot review and quick sequence inspection without licence infrastructure.

DJV 3 has been evolving through incremental releases, culminating for now in version 3.3.3. We missed the earlier 3.x updates, so let’s summarise.

Version 3.3.3 is the latest tagged release. It adds one feature and resolves two bugs. The new feature enables the Escape key to close the secondary window. The fixes address a flipped export issue and incorrect tool tooltips. No further technical detail is provided in the release notes regarding the root cause of the export flip or the tooltip behaviour. Not independently verified at press time. Prebuilt binaries are provided for Linux x86_64, macOS arm64 and Windows amd64. Assets include tar.gz, dmg, exe and zip packages. SHA256 checksums are published for the binaries. To understand the significance of 3.3.3, it is necessary to look at the broader 3.x changes.

Sequence awareness and file handling

One of the more practical changes in the 3.3 branch arrived in version 3.3.0. The built in file browser can now display image sequences as a single item with an associated frame range. Instead of listing every frame as an individual file, the browser aggregates them into a sequence entry.

This change affects directory navigation and review workflows directly. For facilities handling large EXR sequences, reducing file list clutter improves browsing speed and visual clarity. The release notes describe this as a display change in the built in file browser. No additional details are given about sequence detection rules or supported naming conventions beyond the example provided for command line input.

The command line in 3.3.0 now accepts file sequences written in hash notation, such as render.#.exr. This aligns with established VFX naming patterns and allows users to load sequences from the terminal using shorthand notation.

Keyboard shortcuts expanded

Keyboard customisation received significant attention in 3.3.0 and 3.3.2. Version 3.3.0 introduced primary and secondary keyboard shortcuts. It also added a search box to the keyboard shortcut settings. The Tab key can now be assigned to shortcuts, including combinations such as Ctrl+Tab and Ctrl+Shit+Tab for changing the current file. The spelling in the official notes reads Ctrl+Shit+Tab.

A fix in 3.3.2 addressed the keyboard shortcuts search box. Another fix in 3.3.0 resolved keyboard shortcut behaviour in the secondary window. In 3.3.2, the version number was added to the window title, and a command line flag was introduced to print the version. A dialog for displaying system information was also added. These additions are small but relevant for deployment and support scenarios. Being able to query the version from the command line simplifies scripting and diagnostic workflows.

Mouse behaviour and interaction changes

Version 3.3.0 updated mouse configuration in several ways. Extra 1 and Extra 2 buttons were added to the mouse settings. A frame shuttle scale was introduced in the same settings panel. The default mouse mapping was changed so that the left mouse button performs frame shuttle and the middle mouse button pans. This represents a behavioural change for existing users migrating from earlier versions. A separate fix in 3.3.0 corrected swapped middle and right mouse buttons in the settings.

In 3.3.2, pixel and mouse binding labels were added to the colour picker and magnifier tools. The release notes state that both tools now update when picking with the mouse. This suggests synchronised feedback between cursor interaction and tool UI. Version 3.3.3 then added Escape key support for closing the secondary window. Together, these changes reflect ongoing adjustments to interaction ergonomics rather than new feature domains.

Colour management and UI indicators

DJV 3 integrates OCIO for colour management. In 3.3.0, the colour tool was updated to display the OCIO configuration name. At the same time, the path for the OCIO and LUT file names was hidden within the tool.

Status bar indicators were added in 3.3.0 to display whether colour controls, the audio sync offset or the output device are enabled. In 3.3.2, additional status bar indicators were introduced to display when mirroring is enabled.

In 3.3.0, compare overlay and difference modes were fixed to prevent image flipping. In 3.3.3, a flipped export issue was also fixed. The two issues are described separately in their respective release notes, with no explicit statement linking them.

The cumulative effect is a reduction in unexpected image inversion behaviour in both interactive comparison and export contexts, based on the wording of the fixes. The specific code paths affected are not described. Not independently verified at press time.

Playback behaviour

Version 3.3.0 modified playback speed behaviour. The release notes state that playback speed can be increased and decreased by pressing forward or reverse multiple times. This implies incremental speed changes tied to repeated transport input. A magnify tool was added in 3.3.0, along with grid labels. The magnify tool was later enhanced in 3.3.2 with pixel and mouse binding labels and updated mouse interaction behaviour.

Platform specific fixes and build updates

The 3.3.x series includes multiple platform and build fixes.

Version 3.3.0 fixed compilation on Rocky 8 and addressed excessive CPU usage caused by the clipboard on Gnome. It also fixed the mouse settings to only show the Command key on macOS.

Version 3.3.1 was dedicated to macOS package fixes. No further detail is provided in the release notes.

Version 3.3.2 fixed an issue with the window title bar appearing off screen.

The 3.3.0 release notes list a library update to SDL2 release 2.32.10. No additional third party dependency changes are described in the provided notes.

Version 3.3.3 assets include Linux, macOS and Windows builds, with SHA256 hashes published for verification.

What has not changed

Based on the official notes provided, the 3.3 branch does not introduce new file format support, new codec integrations, or changes to core rendering architecture. There is no mention of GPU acceleration updates, threading model changes or cache redesigns in the referenced releases.

There is also no pricing information, as DJV is distributed as free open source software via GitHub.

Any assumptions about performance improvements, stability gains beyond listed fixes, or pipeline scale suitability would be speculative and are not supported by the cited documentation. Not independently verified at press time.

Version 3 so far

Taken together, DJV 3 up to 3.3.3 reflects iterative refinement. Sequence aware browsing, expanded shortcut handling, improved mouse customisation, clearer OCIO labelling, and a series of bug fixes define the branch.

For studios using DJV as a lightweight review tool, the changes are operational. They reduce friction in file browsing, clarify status indicators, and address specific bugs such as flipped compare modes and flipped export.

As always, new tools and updates should be tested in controlled conditions before deployment in production environments.

// https://github.com/grizzlypeak3d/DJV

The post DJV 3.3.3 fixes flipped export and adds Escape first appeared on DIGITAL PRODUCTION and was written by Bela Beier.