Multi-Model Detection Platform

The Truth
Always
Comes Out.

TruthScan is a unified AI-powered platform for detecting deepfakes, AI-generated content, and verifying news against authoritative sources — all in one place.

TRUTHSCAN — ANALYSIS ENGINE
◉ Scanning uploaded image...
→ GAN fingerprint: DETECTED
→ Diffusion artifacts: HIGH
→ Deepfake score: 91%
◉ Running ELA forensics...
→ ELA suspicious px: 38.4%
→ Copy-move: REGION FOUND
→ Originality: 12%
◉ Cross-checking claims...
→ Snopes: DEBUNKED
✗ VERDICT: MANIPULATED + FAKE
🎞️
MODULE 01 · AI & DEEPFAKE DETECTION
AI Content Detector
Upload images, videos, documents, or paste text. Multi-model analysis reveals the percentage of AI involvement using GAN fingerprinting, diffusion classifiers, and NLP perplexity analysis.
Image / Video
Documents
GAN Detect
Deepfake
Open Detector →
📰
MODULE 02 · FAKE NEWS DETECTION
Fake News Verifier
Submit any news claim, article URL, or image. TruthScan cross-references Reuters, BBC, AP, WHO, Snopes, PolitiFact and more to deliver an evidence-backed verdict in seconds.
Claim Extraction
Source Verify
Bias Detection
AI-Powered
Open Verifier →
🔬
MODULE 03 · IMAGE ORIGINALITY
Photo Forgery Detector
Upload any image to detect tampering, splicing, cloning, retouching, and pixel-level manipulation. ELA analysis, EXIF metadata extraction, lighting inconsistency mapping, and GAN forgery detection.
ELA Analysis
EXIF Metadata
Tamper Map
Compare
Open Detector →
🧬
MODULE 04 · CONTENT DNA FINGERPRINT
Content DNA Analyzer
Upload any image, text, or video to generate a unique "DNA fingerprint". Detect duplicates, modified versions, reuploads, and meme evolution using AI embeddings and perceptual hashing.
CLIP Embeddings
pHash
Similarity Search
Duplicate Detect
Open Analyzer →
40+
AI MODELS DETECTED
12
FILE FORMATS
10+
TRUSTED SOURCES
5
FORGERY SIGNALS
4
MODULES
AI Content Detection Platform

Detect
Deepfakes &
AI-Generated
Content

Upload images, videos, documents, presentations, or paste text. TruthScan's multi-model analysis reveals the percentage of AI involvement in any file or content.

🎞️
Drop image or video here
Supports JPG · PNG · WEBP · GIF · MP4 · MOV · AVI · WEBM
📎
📁
Drop document or presentation
Supports PDF · DOCX · DOC · PPTX · PPT · XLSX · TXT · RTF
📎
Analysis takes 2–8 seconds · Results are probabilistic estimates
INITIALIZING ANALYSIS ENGINE...
ANALYSIS COMPLETE — Unnamed
0%
AI CONTENT
How AI Detection Works
01
Upload or Input
Submit any image, video frame, document, presentation, or raw text. TruthScan extracts structured features for downstream analysis.
02
Multi-Model Analysis
Content is processed by specialized detectors — GAN fingerprint detectors, diffusion model classifiers, and large-scale NLP perplexity analyzers.
03
Signal Fusion
Detector outputs are fused using an ensemble meta-model trained on millions of human and AI-generated samples across 40+ AI tools.
04
Confidence Score
A final AI-usage percentage (0–100%) is returned with per-metric breakdowns highlighting which signals drove the verdict.
Supported Content Types
🖼️ JPEG / PNG IMAGE
🎞️ MP4 / MOV VIDEO
📝 Plain Text TEXT
📄 DOCX / DOC WORD
📊 PPTX / PPT SLIDES
📋 PDF PDF
📈 XLSX / CSV SHEET
🌐 Any URL WEB
🎨 WEBP / GIF IMAGE
📺 AVI / WEBM VIDEO
📃 RTF / TXT DOC
🔗 Social Post TEXT
01
EXTRACTING KEY CLAIMS FROM CONTENT
02
DECOMPOSING INTO VERIFIABLE STATEMENTS
03
SEARCHING FACT-CHECK DATABASES
04
CROSS-REFERENCING AUTHORITATIVE SOURCES
05
DETECTING MISINFORMATION PATTERNS
06
COMPUTING CONFIDENCE SCORE
07
GENERATING ANALYSIS REPORT
Fake News Detection Module

Verify
Facts &
Detect
Misinformation

Submit any news article, claim, or image. TruthScan cross-references Reuters, BBC, AP, WHO, Snopes, PolitiFact and more to deliver an evidence-backed verdict.

◈ Submit Content for Verification
⚠ TruthScan will extract and analyze the article text from this URL.
Works with most major news sites, blogs, and public web pages.
🖼️
Drop image or click to browse
JPG · PNG · WEBP · GIF — OCR will extract text for analysis
Preview
Powered by Claude AI + Web Search
Results are probabilistic, not definitive
◈ Verified Source Network
Reuters
INTL NEWS
BBC News
INTL NEWS
AP News
WIRE
WHO
HEALTH
Snopes
FACT-CHECK
PolitiFact
FACT-CHECK
FactCheck.org
FACT-CHECK
PIB India
GOVT
Alt News
FACT-CHECK
UN / CDC
INTL ORG
◈ Verdict Legend
TRUE
Supported by evidence
FAKE
Contradicted by sources
MISLEADING
Partially true, out of context
UNVERIFIED
Insufficient evidence
◈ What We Detect
🔴 Factually incorrect claims
🟡 Misleading statistics or data
🟡 Content taken out of context
⚠️ Sensational / clickbait language
⚠️ Political or ideological bias
⚠️ Missing credible source attribution
📋 Satire misrepresented as news
🤖 AI-generated disinformation
0%
CONFIDENCE
◈ Signal Analysis
◈ Language Flags
◈ Content Analysis — Highlighted
Suspicious sections marked
FALSE / CONTRADICTED
MISLEADING / UNVERIFIED
NEUTRAL / FACTUAL
◈ Claim-by-Claim Breakdown
◈ Evidence & Source References
Report contains full verdict, claims, evidence links, and confidence scores.
⚠ For informational purposes only.
How Fact Verification Works
01
Submit Content
Paste a news claim, article URL, or upload an image. TruthScan extracts the text content for analysis using NLP and OCR.
02
Claim Extraction
Claude AI identifies and isolates individual verifiable claims from the content, breaking it into discrete factual assertions.
03
Source Verification
Each claim is cross-referenced against Reuters, BBC, AP, WHO, Snopes, PolitiFact, PIB and other authoritative sources via live web search.
04
Verdict & Evidence
A final verdict (TRUE / FAKE / MISLEADING / UNVERIFIED) is issued with confidence score, evidence links, and per-claim breakdown.
01
LOADING & DECODING IMAGE DATA
02
RUNNING ERROR LEVEL ANALYSIS (ELA)
03
EXTRACTING EXIF METADATA
04
MAPPING NOISE & COMPRESSION ARTIFACTS
05
ANALYZING LIGHTING & SHADOWS
06
DETECTING COPY-MOVE & SPLICING
07
GAN FORGERY DETECTION
08
COMPUTING ORIGINALITY SCORE
Image Originality & Forensics Module

Expose
Tampering &
Manipulation
Forensically

Upload any photo to detect pixel-level manipulation, copy-move cloning, splicing, retouching, and GAN forgery. ELA heatmaps, EXIF metadata extraction, and lighting inconsistency analysis — all in one scan.

◈ Upload Image(s) for Forensic Analysis
Primary Image
CHANGE IMAGE
📸
Drop image here
JPG · PNG · WEBP · BMP · TIFF
Real canvas-based ELA + Claude AI analysis
Supports JPG, PNG, WEBP up to 20MB
0%
ORIGINAL
◈ Error Level Analysis (ELA) — Heatmap
Bright = high error = possible manipulation
ORIGINAL
Original
ELA OUTPUT
ELA · AMPLIFIED ×10
LOW COMPRESSION ERROR (likely original)
MEDIUM ERROR (re-saved / edited region)
HIGH ERROR (suspected manipulation)
◈ Tamper Heatmap — Suspicious Regions
Red overlay = suspected manipulation zone
Tamper map
HIGH SUSPICION — Copy-move / splice
MEDIUM SUSPICION — Retouch / filter
CLEAN REGION
◈ Forensic Signals
◈ Key Findings
◈ EXIF Metadata Analysis
Forensic report includes originality score, ELA data, EXIF metadata, and all findings.
⚠ Probabilistic results — not admissible as sole legal evidence.
How Image Forensics Works
01
ELA — Error Level Analysis
Re-encodes the image at a fixed quality and amplifies the difference. Edited regions have different compression error levels than untouched areas, revealing manipulation zones.
02
EXIF Metadata Extraction
Reads embedded camera metadata — device model, GPS, timestamps, software. Flags suspicious entries like editing software fingerprints, mismatched timestamps, or stripped metadata.
03
Noise & Compression Analysis
Detects inconsistent noise grain and JPEG block artifacts across image regions. Spliced or cloned areas typically show different noise characteristics than the native image.
04
Lighting & GAN Forgery
Analyzes shadow directions, highlight coherence, and color temperature consistency. GAN detection flags frequency-domain artifacts and pixel-level patterns unique to AI-generated content.
Content DNA Fingerprinting System

Content
DNA
Analyzer.

Generate unique "DNA fingerprints" for any content. Detect duplicates, modified versions, reuploads, and meme evolution using CLIP embeddings, Sentence-Transformers, and perceptual hashing.

◈ Content Input
🧬
Drop an image here
JPG, PNG, WEBP — Max 20MB
📎
🎬
Drop a video here
MP4, WEBM, AVI — Max 50MB
🎬
Search generates DNA fingerprint and compares against stored content database
1
Preprocessing content...
2
Generating CLIP embeddings...
3
Computing perceptual hash...
4
Searching vector database...
5
Classifying similarity...
0%
SIMILARITY
◈ Similarity Breakdown
◈ Fingerprint Details
How Content DNA Works
01
Upload Content
Upload an image, paste text, or submit a video. The system extracts features for fingerprinting regardless of format or quality.
02
Generate Embedding
CLIP generates 768-dim image vectors; Sentence-Transformers produce 384-dim text vectors. A perceptual hash captures pixel-level features.
03
Vector Search
The embedding is compared against all stored fingerprints via FAISS cosine similarity. Combined with pHash distance for images for accuracy.
04
Classification
>90% = Duplicate, 70-90% = Modified, 50-70% = Similar, <50% = Original. Results include top matches, scores, and metadata.
Audio Authenticity & AI Voice Detection

Vocal
Truth
Detector.

Analyze audio recordings to detect AI voice cloning, deepfake speech, and temporal tampering using spectral discontinuity and prosody analysis.

🎧
Drop audio file here
WAV, MP3, AAC, OGG — Max 25MB
◈ Signal Waveform
ORIGINAL
98%
CONFIDENCE
🤖 AI Voice Analysis
AI Likelihood 2%
Synthetic Prosody 0.12
Phoneme Consistency 0.05
Detected Model None
✂️ Tampering Analysis
Editing Likelihood 5%
Spectral Discontinuity 0.08
Phase Inconsistency 0.15
Noise Floor Mismatch NO
◈ Suspicious Segments
No significant anomalies detected in signal timeline.
Summary Explanation

Platform Architecture

Built for
Truth &
Transparency

TruthScan is a two-module platform combining probabilistic AI detection with real-time fact verification powered by Claude AI and live web search.

Platform Modules
01
AI & Deepfake Detector
Simulated multi-model pipeline analyzing GAN fingerprints, diffusion artifacts, perplexity, and burstiness across images, video, text, and documents.
02
Fake News Verifier
Real AI-powered module using Claude Sonnet + live web search to extract claims, verify against 10+ authoritative sources, and return structured verdicts.
03
Image Originality Detector
Real canvas-based ELA using browser APIs. Claude AI analyzes pixel statistics, EXIF metadata, and noise patterns to detect tampering, splicing, copy-move, and GAN forgery.
04
Ethics & Transparency
Results are clearly labeled as probabilistic estimates. The system never fabricates sources, always shows evidence, and returns "Inconclusive" when data is insufficient.
Source Network
📡 Reuters NEWS
📻 BBC News NEWS
📰 AP News WIRE
🌍 WHO HEALTH
🔍 Snopes FACT-CHECK
⚖️ PolitiFact FACT-CHECK
✅ FactCheck.org FACT-CHECK
🇮🇳 PIB India GOVT
🗞️ Alt News FACT-CHECK
🌐 UN / CDC INTL ORG
⚠️ TruthScan provides probabilistic analysis. Results are not 100% conclusive and should not be used as sole evidence in legal, academic, or official proceedings.
Fake News module powered by Claude AI + live web search. AI Detection module uses simulated analysis. Neither replaces professional journalism or legal review.