Adobe and NVIDIA expand AI work across Photoshop, Premiere and Frame.io

Adobe and NVIDIA announced an expanded collaboration to accelerate AI and media workflows in Photoshop, Premiere, Substance 3D, Frame.io and Brand Concierge using NVIDIA hardware, software libraries and models.

The work spans local creative applications, Frame.io’s cloud processing pipeline and an enterprise customer-experience product. Availability differs across the creative tools: an NVIDIA companion post says Photoshop and Premiere AI feature optimizations are ready in beta, while Substance 3D updates are coming soon. It does not give general-availability dates for those optimized builds.

In Photoshop, Adobe reworked compositing around graphics processor acceleration, the companies said. Examples include the Remove Tool and on-canvas effects such as Gaussian Blur and Light Leaks. The Premiere work covers playback, rendering, effects, color and AI features including Enhance Speech and Object Mask. The companies also said RTX Spark accelerates demanding texturing steps in Substance 3D Painter.

Adobe and NVIDIA claim the RTX Spark optimizations can make AI processing, editing, color and effects up to twice as fast across the three applications. The October 15 announcement did not disclose the hardware configurations, software builds, baselines, datasets or test methodology behind that figure. No independent reproduction of the result was located during research.

This creative-application work centers on RTX Spark for local AI workflows, unlike data-center deployments such as CoreWeave’s Vera Rubin cloud rollout.

Frame.io is receiving a separate infrastructure change. Adobe said the updated pipeline shifts video decoding, preprocessing and the creation of searchable numerical representations called embeddings to NVIDIA GPUs. It uses NVIDIA’s NVDEC for hardware-assisted video decoding and TensorRT with CUDA for preprocessing and embedding generation.

For selected Frame.io transcode and search workloads, the companies reported roughly five times faster embedding generation, a four-to-five-times improvement in 95th-percentile processing latency — the latency threshold below which 95% of measured runs fall — and a search proxy about 80% smaller. These are also company figures, with no published test configurations or independent validation. The announcement did not say whether the updated pipeline is available to all customers or identify eligible plans, regions or a rollout date.

The collaboration extends to Brand Concierge, Adobe’s enterprise software for personalized conversations grounded in an organization’s information. Adobe and NVIDIA said they optimized a version with NVIDIA Nemotron open models and plan to make it available to enterprise customers later in 2026. NVIDIA has selected Brand Concierge for use on NVIDIA.com, but the companies did not give a deployment date or say that the experience is already live.

The expansion follows the companies’ March strategic partnership, which covered Firefly models, agent-based creative and marketing workflows, 3D digital twins, Frame.io cloud acceleration and broader use of NVIDIA infrastructure, libraries and models across Adobe products. Adobe and NVIDIA did not specify when Brand Concierge will appear on NVIDIA.com.