Open-source deepfake detection: train your own model with TensorFlow, Keras & EfficientNet
Agent skill for deepfake detection & media safety — detect AI-generated audio, images, and video with Resemble AI
A simple library that implements CLIP guided loss in PyTorch.
🎬 Automated AI avatar news video generator built with n8n. Fetches RSS headlines, generates scripts with Google Gemini, and creates professional videos with HeyGen's AI avatars. Fully automated from feed to final video.
Face Animation from Text 🧙♂️
PIN Architecture: a six-layer forensic system for detecting AI-generated and manipulated imagery. Fifteen independent analysis pins run concurrently: C2PA provenance, compression forensics, CLIP/SigLIP2/frequency detectors, Grad-CAM explainability, LLM adjudication and a calibrated XGBoost ensemble. Signed provenance outranks statistical inference.
AI video studio on the ibl.ai platform for generating avatar videos, cloning voices, and authoring scripts from a single workspace. Built with Next.js, TypeScript, and Tailwind CSS on the @iblai/iblai-js SDK, with a hosted backend at iblai.app. Runs on desktop and mobile.
A public license for persona, identity, and synthetic-media reuse.
Open Source Deepfake Detection Models in One Place
Official repository for AADD-2025: 1st Adversarial Attacks on Deepfake Detectors Challenge - ACM Multimedia 2025 Grand Challenge. Investigating adversarial vulnerabilities in deepfake detection models through adversarial perturbations that evade classifiers while preserving visual similarity.
Aletheia is a multimodal ML architecture for forensic detection of synthetic video. Integrating deep learning and NLP, it analyzes spatiotemporal anomalies, biometric artifacts, and audio-linguistic misalignments. The framework performs cross-modal feature extraction to provide deterministic probability scoring for media authentication.
Source media and persona reference material for Skye under the Open Persona License.
Expert survey on GenAI disinformation threats & countermeasures - instrument, data, and publications. Published in HKS Misinformation Review (doi:10.37016/mr-2020-205) and at R2CASS @ WWW 2026. Survey still open for expert participants.
Real-time deepfake detection Android app using multi-modal AI fusion. Features CNN inference, artifact analysis & on-device ML. Built with Kotlin, Jetpack Compose & TensorFlow Lite.
DeepFake detection with Deep Tree Network (DTN) architecture - 94.5% accuracy, 45 FPS real-time processing with zero-shot learning
Analysis of 13 years of BBC articles (2012–2025) using the isFake.ai AI text detector.
Privacy-first image authenticity workbench combining C2PA provenance, digital forensics, metadata analysis, and local AI-generated image screening with React, FastAPI, and ONNX.
🎬 Automated portrait (9:16) video generator for TikTok, Reels & Shorts. Powered by local Ollama LLMs, Edge-TTS, GPU-accelerated local RVC V2 voice conversion, Dynamic subtitles, and MoviePy smart B-roll stitching. 100% offline-ready with an elegant PySide6 Desktop GUI. 🚀
Decentralized detection network against synthetic media and deepfakes.
AI-powered voice deepfake detection platform combining ensemble ML, anomaly detection, and explainable AI to strengthen cybersecurity resilience and responsible AI governance.
To associate your repository with the synthetic-media topic, visit your repo's landing page and select "manage topics."