Ensemble speaker verification achieving 97% accuracy - Intelligent fusion of MFCC+DTW (92%) and Resemblyzer CNN (94%) for voice authentication

Audio analysis toolkit

Desktop GUI tool for detecting audio deepfakes using hand-crafted features + XGBoost/LightGBM ensemble. Segment-level analysis, heatmaps, PDF reports. Zero deep learning required.

Deepfake voice detection web app. Extracts 7 features and uses an LLM forensic analyzer to flag deepfakes. Next.js + Supabase.

Text2Speech2Clone Data Set

Free, local-first and explainable AI-origin music analysis for artists, labels, platforms, rights teams and researchers.

A robust AI-powered web app to detect deepfake audio using Wav2Vec2 and FastAPI. Deployed on Hugging Face Spaces for the CyberSecurity Hackathon.

A web-based Audio Forensic Intelligence Platform for audio authenticity verification, fingerprint generation, tamper detection, watermark validation, and forensic report generation using Django and Librosa.

Forensic toolkit for detecting AI-synthetic voices through Nonlinear Phenomena (NLP) and phase-space reconstruction.

Detect fake FLAC files — find MP3/AAC/Opus transcodes disguised as lossless audio. Fast Rust CLI + web UI.

A Decoupled Tri-Modal Architecture for Real-Time Deepfake Interception via WebRTC and Late-Fusion.

Explainable deepfake voice detection — RawNet2 + Whisper + real-time browser inference. 95.2% accuracy, 4.19% EER on ASVspoof 2019.

AI voice security platform detecting synthetic speech & voice cloning attacks, fusing WavLM transformer embeddings + LFCC forensics + LightGBM — 99.84% accuracy.

a tiny python tool to generate spectrograms for audio files and conduct heuristic "lossy source" analysis.

Explainable deepfake audio detection using CNN-LSTM, LFCC, and GradCAM.

Local deepfake voice detection with Wav2Vec2 and acoustic liveness checks.

Full-stack forensic speech analysis platform for detecting AI-generated vs human audio with real-time inference and explainable ML.

A FLAC steganography toolkit for encrypted payloads, audio forensics, and PCM embedding benchmarks.

Forensic Audio Classifier Tool is an ML-based digital forensics system built using PyTorch, Transformers, and a custom hybrid pipeline (Acoustic Model + Language Model + Classifier). It is designed for the Tripura Bengali dialect, enabling accurate transcription, keyword detection, and automated (Flagged / Review / Safe) audio classification.

Deepfake audio detection thesis: ML & deep learning models with live benchmark & Streamlit demo