Summary
Adobe and NVIDIA say they are optimising Photoshop, Premiere and Substance 3D for NVIDIA’s RTX Spark, with claimed gains of up to 2x in several tasks. The expanded work also targets Frame.io’s cloud media search and Adobe Brand Concierge.
Adobe and NVIDIA announced an expanded collaboration on 15 October 2026, covering creative applications, cloud video workflows and AI-powered customer interactions. The companies say optimisations for NVIDIA’s RTX Spark will bring up to 2x faster AI processing, editing, colour and effects in Photoshop, Premiere and Substance 3D. RTX Spark is scheduled to launch on 16 October.
The announcement also describes changes to Frame.io’s cloud processing pipeline and an Adobe Brand Concierge version optimised with NVIDIA Nemotron open models. The work extends a strategic partnership announced in March, which included GPU acceleration for Frame.io and projects involving Substance 3D, Firefly Foundry and enterprise 3D digital twins.
RTX Spark optimisations target specific creative tasks
Adobe says it rearchitected Photoshop around GPU-accelerated compositing. The company points to tasks such as removing distractions with the Remove Tool and applying Gaussian Blur or Light Leaks directly on the canvas as examples of work intended to feel faster and more responsive.
In Premiere, the companies say RTX acceleration covers demanding post-production work, including playback, rendering, effects, colour and AI-powered tools. They cite Enhance Speech and Object Mask as features expected to benefit. For Substance 3D Painter, the stated focus is speeding up resource-intensive texturing steps.
The tools are being optimised for RTX Spark’s unified memory and AI computing. Unified memory is a shared pool accessible to different parts of a computing system; in principle, it can reduce the need to move data between separate memory pools. The announcement does not detail how each application uses that design, so the reported speed gains should be read as company claims about the optimised workflows, not as a single performance result applying equally to every task.
Frame.io processing and Brand Concierge
Adobe is updating Frame.io’s cloud pipeline to use NVIDIA video acceleration for media decoding, preprocessing and content-embedding generation. Decoding converts compressed video into frames that can be processed. Embeddings are numerical representations that software can use to organise or search content by its characteristics.
The pipeline uses NVIDIA NVDEC for hardware-accelerated video decoding and NVIDIA TensorRT with CUDA for preprocessing and generating embeddings. NVIDIA reports early results from selected Frame.io transcoding and search workloads: approximately 5x faster embedding generation, a 4–5x improvement in 95th-percentile processing latency, and a search proxy approximately 80% smaller. The 95th percentile refers to the slower end of the measured workload: 95% of measured processing times were at or below that point. The companies say faster processing can make uploaded content searchable sooner and help large media libraries refresh more efficiently.
For customer-facing AI, Adobe and NVIDIA have optimised Brand Concierge with NVIDIA Nemotron open models. The version is expected to be available to enterprise customers later in 2026. NVIDIA says it has also selected Brand Concierge for personalised interactions on NVIDIA.com, grounded in the company’s organisational data and content. The announcement describes this as a way to tailor conversations to customer needs while using Nemotron for responses at scale.