TL;DR:
- The Apple M2 Studio offers exceptional speed and is tailored for VFX Machine Learning.
- The integration of Foundry’s NUKE with Machine Learning and CopyCat runs smoothly on M2 Silicon.
- The compact and quiet design of the Mac Studio is perfect for collaborative VFX environments.
- CopyCat, introduced in Nuke 13, is a game changer for artists, saving time and enhancing workflow.
- The new Nuke version running on Apple Silicon is impressively fast and outperforms high-end PCs.
- The Foundry’s focus on Apple Silicon is driven by the demand from VFX supervisors and ML’s transformative impact.
- Apple’s commitment to providing robust ML tools is exemplified by Nuke’s implementation on Apple Silicon.
- The Apple M2 Studio’s success demonstrates the power and appeal of Apple Silicon in the VFX market.
Main AI News:
In the dynamic world of visual effects (VFX), innovation is a driving force that propels professionals to achieve new heights. Enter the Apple M2 Ultra Studio, a groundbreaking tool that has recently taken the industry by storm. At the fxguide tech lab, we were fortunate to experience the power and speed of this cutting-edge workstation firsthand, and we can confidently say that every VFX professional should pay attention to its remarkable capabilities.
Speed That Leaves You Breathless
The first standout feature of the Studio M2 Ultra is its exceptional speed. When it comes to VFX Machine Learning, this workstation leaves its competitors in the dust. The integration of Foundry’s NUKE, enhanced with Machine Learning and CopyCat, harnesses the M2 Silicon to deliver unprecedented performance on the Apple M2 Studio. The sheer velocity at which it operates is truly awe-inspiring.
Unveiling the Studio M2 Ultra
During the WWDC launch event, fxguide had the privilege of extensively testing the Mac Studio, primarily focusing on its application in Machine Learning (ML). As a publication deeply immersed in the world of ML, we are thrilled to share that the upcoming Nuke release will be fully optimized for Apple Silicon, boasting PyTorch support that unlocks the astounding potential of CopyCat ML functionality.
The Power of Desktop VFX Applications
In the era of cloud computing, it’s easy to overlook the advantages of powerful desktop applications. While remote cloud rendering and other cloud-based tools have their merits, certain VFX applications thrive when executed directly on your desktop. Take, for instance, VFX editing for 4K UHD feature films, which demands uninterrupted playback and seamless reviewing without the pitfalls of delayed uploads or compromised compression.
The Apple Studio, nearly identical to the Mac Pro, offers the same VFX editing and ML capabilities. Our tests with the M2 Ultra, equipped with a 24-core CPU, 76-core GPU, and 32-core Neural engine, accompanied by 192 GB of unified memory, mirrored the performance of the Mac Pro tower configuration. Moreover, the Studio’s compact design and whisper-quiet operation make it an inconspicuous powerhouse, ideal for collaborative VFX environments.
A Leap Forward in Machine Learning
While Autodesk Flame and Maya have already embraced Apple Silicon natively, Nuke has only recently made its way to Apple Silicon computers, relying solely on CPU-based processing. However, the advent of CopyCat by The Foundry has completely transformed the game. Introduced with Nuke 13 and further expanded in version 13.2, CopyCat revolutionizes artists’ ability to harness the immense power of machine learning.
By providing example frames and source frames for comparison, CopyCat learns from a small dataset to accurately convert images according to an artist’s vision. The subsequent Inference node extrapolates this understanding across an entire clip, offering unprecedented efficiency. Nuke 14 introduced Cattery, broadening the range of applications that artists can explore.
CopyCat’s potential spans various domains, from enhancing roto and compositing to rigging, cleanup, and beauty work. With features like de-blurring, smart vector generation, segmentation, and real-time test comps, the possibilities are virtually limitless. CopyCat empowers experienced artists by streamlining VFX workflows. Unlike opaque AI algorithms, CopyCat provides artists with a controllable, repeatable, artist-driven, understandable, and explainable AI implementation, ushering in a new era of VFX possibilities.
Unleashing the Potential
Imagine taking the mesmerizing “style transfer” demonstration featuring Ballerina and effortlessly running it on the new Nuke using the Mac Studio. With the right file, CopyCat seamlessly executes the desired transformation without any additional requirements. The power lies in the simplicity of the process. The new Nuke version, still in development and not yet in beta, exhibits remarkable speed, surpassing the capabilities of our high-end PCs equipped with top-tier graphics cards.
It’s worth noting that this improved performance owes its success not only to Apple’s hardware but also to The Foundry’s internal enhancements to CopyCat. As this new Nuke version enters private beta soon, we anticipate an array of exciting features to be unveiled, and our informed speculation suggests that SIGGRAPH in Los Angeles will serve as the ideal stage for showcasing the upcoming NUKE release in all its glory.
The Foundry’s Focus on Apple Silicon: A Strategic Move
Some may wonder why The Foundry devoted substantial effort to optimizing Nuke for Apple Silicon, considering that some diehard PC users remain loyal to their preferred platform. The answer lies in the transformative impact of machine learning on the industry. VFX supervisors and on-set crews have consistently voiced their desire for Nuke to embrace Apple Silicon. A MacBook Pro, even before the latest M2 upgrades, has become the laptop of choice for most VFX supervisors. However, there’s more to it than personal preferences.
Unlocking the Power of CopyCat
The concept of CopyCat inference encourages supervisors to train inference nodes for sequences rather than individual shots. Training on a set of approximately 25 images enables the inference node to be used across all similar shots within the sequence. Mastering the art of selecting the appropriate training data is pivotal in achieving consistent results.
While we have yet to conduct thorough testing, The Foundry’s development team successfully ran a full CopyCat node on an older M1 MacBook Pro, proving its compatibility and efficacy. Although an M1 may lack the graphics RAM to compete with the M2 Ultra, Nuke seamlessly operates on the base M1 Pro Apple hardware, an ideal choice for supervisors when assigning shots to their team. While the M2 Ultra Studio stands as the flagship option, other viable alternatives ensure a smooth transition to the lightning-fast Apple Silicon architecture.
Apple’s Commitment to VFX
Although VFX might not be Apple’s primary market, its dedication to advancing machine learning is undeniable. Apple aims to provide robust tools for ML, and the implementation of Nuke on Apple Silicon exemplifies their commitment to opening up their hardware to a wider audience.
Breaking Free from Limitations
Until recently, ML applications predominantly relied on NVIDIA hardware due to its CUDA platform. NVIDIA has played a crucial role in fostering ML innovation. However, their CUDA GPU application programming interface presented a limitation, constraining certain GPU-based ML processes to their hardware. When the Mac Pro was initially released, we highlighted the lack of NVIDIA CUDA compatibility as a major drawback, restricting Apple’s suitability for a range of ML applications.
A Shifting Landscape
With the advent of Apple Silicon, the CUDA conundrum has transformed from a solitary solution to one of many possibilities. In this regard, healthy competition benefits both Apple and NVIDIA. Apple’s Silicon implementation balances power efficiency and heat output, delivering unparalleled GPU performance in a compact form factor. While plugin cards offer flexibility, the sleek, fast, and silent nature of the Mac Studio’s out-of-the-box functionality holds significant appeal for VFX professionals.
Machine learning applications deserve a dedicated space on your desk, providing direct control without relying on cloud services. Apple has undoubtedly achieved this feat, as evidenced by the remarkable success of the fully functional Nuke CopyCat solution. Critics may dismiss it as a niche attraction for Apple enthusiasts, but the Apple Studio M2 Ultra’s exceptional performance substantiates its position as an outstanding tool in the VFX industry.
Conclusion:
The introduction of the Apple M2 Studio, with its exceptional performance and optimized support for VFX Machine Learning, represents a significant milestone in the VFX industry. The seamless integration of Foundry’s NUKE with CopyCat ML functionality on Apple Silicon opens up new avenues for artists to harness the power of machine learning, enhancing their workflows and unlocking unprecedented creative possibilities. This advancement not only caters to the demands of VFX professionals but also positions Apple as a key player in the machine learning market, solidifying its commitment to delivering powerful, efficient, and user-friendly hardware for a wide range of industries.
