NVIDIA unveiled a synthetic video detector for newsrooms and broadcasters at SIGGRAPH 2026, while highlighting how creative tools from Adobe to Unreal Engine are opening up to AI agents via MCP.
Busy time for NVIDIA at SIGGRAPH last week. The company doubled down on the AI that has done so much for its fortunes, while trying to rein in the epidemic of fake news its technology has also fuelled.
In that latter category, its new Synthetic Video Detector NIM microservice, part of its AI for Media platform, hopes to give newsrooms and broadcasters an AI-assisted signal for spotting synthetic video.
The microservice analyzes video frame by frame, producing a classifier score for whether a clip contains synthetic content. Editorial teams can use that score to prioritize footage for review, flag or quarantine questionable clips, or escalate them for further checks.
NVIDIA positions it as an additional signal for verification workflows, not a replacement for existing fact-checking practices. Synthetic Video Detector is designed to hold up after the compression, resizing, cropping and re-encoding common in newsroom and social video pipelines.
In the company’s own testing, accuracy reached up to 92% on uncompressed video, 87% at 15% compression and 82% at 50% compression.
Processing time for 1080p video is as little as 22 milliseconds on NVIDIA RTX systems and around 30 milliseconds on NVIDIA L40 GPUs. It can be deployed on-premises, at the edge, in hybrid setups or in approved air-gapped environments. Wowza is the first announced partner, embedding the microservice into its Video Intelligence Framework to bring real-time synthetic video detection to livestreaming workflows across more than 35,000 deployments in over 170 countries.
The NVIDIA Synthetic Video Detector NIM microservice is available now at build.nvidia.com.
That list starts with Adobe, which as we’ve already covered, is expanding its creative agent across Firefly, Express and Creative Cloud, letting users describe a desired outcome while the assistant orchestrates multistep workflows. Adobe is also bringing its creative tools to third-party AI platforms through the Adobe connector, and offers an Adobe Express Developer MCP Server for building Express add-ons.
Affinity by Canva has added an AI Connector for Claude, using MCP to let designers ask Claude to rename layers and artboards, resize and reformat assets for multiple channels, apply bulk edits, optimize vector paths and prepare files for delivery. Claude can also help build reusable scripts and custom features.
Blender offers a lightweight MCP server through Blender Lab, giving a natural-language interface to Blender's Python API, documentation and complex setups.
Boris FX Silhouette now includes an MCP server that exposes its FX Scripting API as MCP tools, letting AI assistants inspect projects, build node trees, edit shapes and keyframes and render frames. A new preferences panel installs the MCP package, generates a client configuration and tests the connection. Interactive online mode connects to an active session; offline mode runs headless instances for automation and batch processing.
Foundry Griptape natively supports MCP for VFX pipelines, letting studios manage multiple AI models and agents with traceability and creative control. Integrated with tools like Blender and Foundry Nuke, it can handle cleanup, matte painting and quality control while artists retain final say.
SideFX is bringing MCP support to Houdini 22 through its new APEX Script workflow, giving AI assistants access to APEX Script syntax, functions, documentation and examples to help generate and refine procedural character rigs. Community-developed MCP servers offer broader ways to interact with Houdini.
And one of the biggest beasts of all in the VFX world, Unreal Engine, now lets AI clients connect to Unreal Editor through MCP, enabling agents that can reason over scenes, assets and project state.