ClouderaNOW Activate your data, unleash your AI, and scale across any cloud | Dec 9

 Register
    • Cloudera Cloudera
    • Contact us(888) 789-1488
    International:+1 (650) 362-0488
    Email Sales
    Support Portal
    AI Blueprints AI Blueprints
    • Blueprint Card

    Overview

    Based on the Nvidia Blueprint, the Altered Video Detector blueprint helps teams assess clip-level synthetic likelihood in MP4 media inside a Cloudera AI project. A Launchpad application configures deployment, builds the UI, and generates standalone runtime applications. Bundled mode runs open-weights models (VideoMAE or frame classifiers) on a GPU in your tenant—no external inference API required for detection. Serverless mode sends video to NVIDIA-hosted Synthetic Video Detector inference over gRPC when you supply an NGC API key and NVCF function ID. Both paths share one custom runtime image edition (SyntheticVideoDetector) and the same Next.js detection experience.

    Demo

    This demo analyzes two sample videos, one authentic and one deepfake, to determine whether each clip is real or synthetic. The demo takes the user through two different journeys: using Nvidia’s API-hosted model as well as using an offline HuggingFace model. Watch the demo

    Key Features

    • Flexible deployment options: run locally using GPU resources or directly via NVIDIA cloud services.

    • Built-in Launchpad app to configure settings, launch runtime applications, and monitor server status.

    • Users can add their own detection models via a HuggingFace integration or select from one of the sample ones in the Launchpad.

    • Interactive dashboard displaying frame-by-frame timelines and overall synthetic probability scores.

    • Pre-built environment configuration for easy registration in your Runtime Catalog.

    • Supports both self-hosted open model weights and cloud-hosted API execution.

    Use Case

    Problem: Newsrooms, trust & safety teams, compliance officers, and security analysts face an increasing volume of synthetic and AI-manipulated media. Distinguishing authentic footage from deepfakes requires rapid, reliable, and standardized clip-level verification to prevent misinterpretation, reputation damage, and fraudulent operations.

    Outcome: Organizations deploy a fully governed Cloudera AI application that processes MP4 video inputs, calculates frame- and clip-level synthetic likelihood scores, and delivers actionable detection reports. Teams can leverage Bundled mode for secure, air-gapped GPU inference using HuggingFace weights within their local environment, or utilize Serverless mode to connect directly with NVIDIA-calibrated cloud APIs for high-throughput production validation.

    Architecture

    The architecture connects a Cloudera AI project running a Launchpad application to either a Serverless or Bundled detection runtime. In Serverless mode, the CPU pod sends requests to the Detect/Demo UI, which routes via TLS and an NGC key to NVIDIA NVCF gRPC services. In Bundled mode, the GPU pod runs a local HF open server alongside the UI to perform local inference over HTTP.

    Architecture diagram for Altered Video Detection

    Get started by deploying this blueprint in your environment

    Every blueprint ships with source code, sample data, and step-by-step instructions — ready to run on your Cloudera Platform

    Your form submission has failed.

    This may have been caused by one of the following:

    • Your request timed out
    • A plugin/browser extension blocked the submission. If you have an ad blocking plugin please disable it and close this message to reload the page.