Industry Marketing Manager - AI

Foundry launches Griptape Enterprise, putting governed AI pipelines inside the studio firewall. Jul 16, 2026 Foundry launched Griptape Enterprise, a new license tier of its Griptape AI workflow orchestration platform built to run generative pipelines entirely inside a studio's own infrastructure. The tier is available now. * On-premises or private-cloud deployment keeps plates, models, and metadata behind the studio firewall. * Modular diffusion pipelines run FLUX, Stable Diffusion XL, LTX, and WAN models locally on studio GPUs. * Kyle Roche, Griptape co-founder, steps into a new role as Foundry's Chief AI Officer. Foundry positions it, in its launch announcement, as a governed alternative to the improvised setups studios have been building to bolt AI models onto existing pipelines. A studio-controlled license tier aimed at security and governance. Griptape Enterprise sits on top of the base Griptape platform and adds the controls high-end facilities need before they let generative tools touch client footage. It offers studio-controlled licensing and permissions, plus enterprise governance with centralized, unlimited licensing across a facility. Deployment runs on-premises or in a private cloud, so proprietary plates and trained models never leave studio-owned hardware. Color handling uses OpenColorIO for studio-grade plate and color management, and show-ready projects ship with templates for workspaces and versioning. VP Land covered Foundry's acquisition of Griptape, the deal that brought the orchestration technology in-house and set up this enterprise tier. Nuke integration publishes workflows as versioned gizmos. Griptape Enterprise connects directly to Nuke. Artists can publish workflows as versioned gizmos and run scripts headlessly, which pushes generative steps into the same versioned, repeatable structure compositors already use for the rest of a shot. The tier also adds agentic automation with a choice of local or cloud LLMs, and it ships an MCP server that lets AI agents pilot Griptape at scale. VP Land previously detailed how Foundry brought Griptape agents into Nuke, Blender, and Maya through MCP, the same protocol underneath this release. Modular diffusion pipelines run open models on studio GPUs. The generative core is a set of modular diffusion pipelines that support FLUX, Stable Diffusion XL, LTX, and WAN models running locally on studio GPUs. Keeping inference on local hardware matters for facilities that cannot send frames to an outside API for legal, security, or client-confidentiality reasons. That local-first design lines up with Foundry's broader push to fold machine learning into its compositor, which VP Land saw in Nuke 17.0 and its expanded machine-learning toolset. Kyle Roche moves from Griptape CEO to Foundry Chief AI Officer. Kyle Roche, Griptape co-founder and its former CEO, has been named Foundry's Chief AI Officer. He joined the company in February 2026 when the acquisition completed. Roche framed the enterprise tier as a way to move studios into AI production without the one-off workarounds many have assembled on their own: Foundry has been a trusted technology partner to visual effects and animation studios for over 30 years. With the new Griptape Enterprise license tier, Foundry is shepherding high-end creative teams into AI-enhanced production in a safe and secure way, with artistry and craft at its core. While studios are navigating how best to integrate AI models with bespoke platforms and workarounds, VP Land is pleased to offer Griptape as a highly customizable and safe off-the-shelf alternative. Kyle Roche, Co-founder and Chief AI Officer What governed, off-the-shelf AI means for studio pipelines. For facilities weighing generative tools, the constraint has rarely been model quality. It has been control: where the data lives, who holds the license, and whether an AI step can slot into a versioned pipeline without breaking review and audit trails. Griptape Enterprise answers those questions by keeping deployment, models, and color management inside the studio's own walls, and by wiring the whole thing into Nuke's existing gizmo and scripting workflow. Whether high-end studios adopt a packaged orchestration layer over their own internal tooling is the open question. With FLUX, Stable Diffusion XL, LTX, and WAN all running locally and Nuke as the front end, Foundry is betting that the facilities most cautious about AI are the ones most likely to pay for a governed way in.

Foundry ships SmartRoto for Nuke, promising Up to 4x faster rotoscoping. Jul 16, 2026 Foundry has released SmartRoto, an AI rotoscoping add-on for Nuke that fills in the frames between an artist's keyframes. According to Foundry, the tool moves and deforms splines across a shot from a few input frames, while leaving the artist in control of the final shape. * Up to 4x faster. Foundry says SmartRoto can cut rotoscoping time to roughly a quarter of a manual pass. * Runs locally, trained on licensed data. The model processes shots on the artist's own machine, and Foundry says it was trained on licensed footage. * $499 to start. The introductory licence is priced at $499 and rises to $599, with a 90-day trial that closes August 16, 2026. How SmartRoto handles the in-between frames. Rotoscoping is the manual work of drawing and animating mattes around objects so compositors can isolate them, and it remains one of the most time-intensive tasks in a VFX pipeline. SmartRoto targets the slowest part of that job: the tweening between an artist's keyframes. The artist still sets up the initial splines and a small number of keyframes. SmartRoto then predicts the intermediate shapes across the sequence, tracking and deforming the splines to follow motion in the plate. Foundry says the artist keeps control throughout and can correct any frame, so the tool assists the roto pass rather than replacing the artist's judgment on tricky edges. From a DNEG research project to a shipping product. SmartRoto did not start as a product. It began as a Foundry research collaboration with visual effects studio DNEG and the University of Bath, documented in Foundry's own research write-up. The model was trained on DNEG's production rotoscopy artwork, a dataset the team describes as more than 650,000 artist-animated shapes and 125 million keyframes drawn from real shot work. The research goal was more modest than the launch claim. Ben Kent, a research engineering manager at Foundry, framed the original target as saving "even 25% of an artist's time." The commercial release now advertises up to a 4x speedup, a larger figure that VFX teams will want to test against their own footage before budgeting around it. Part of Foundry's longer AI build-out in Nuke. SmartRoto lands on top of years of machine learning work inside Nuke. Foundry's AI lineage in the compositor traces back to CopyCat, and the company has kept expanding those capabilities. VP Land covered Nuke 17.0, which added native Gaussian Splat support and broadened its machine learning toolset. The company has also been buying and building around AI infrastructure. VP Land reported on Foundry's Griptape acquisition, which brought Python-based agent orchestration into its pipeline tools. SmartRoto fits the same pattern: targeted AI applied to a specific, repetitive artist task rather than a general-purpose generator. AI roto joins a crowded problem space. SmartRoto enters a category where other teams are also attacking hard matte work. VP Land previously covered CorridorKey, the open-source neural keyer that tackles green screen edges many traditional keyers struggle with. Where CorridorKey focuses on keying, SmartRoto focuses on the spline-based roto that artists fall back on when keying is not an option. For studios weighing it, the practical questions are cost and fit. At an introductory $499 per licence and a local install that keeps footage off external servers, SmartRoto is priced for individual artists and small teams as much as large facilities. The 90-day trial gives roto artists a window to run it against their own shots before the price steps up to $599, and before the August 16, 2026 deadline closes.