Media Provenance API
Trace the origin of images and video before your agents act on them. Verify C2PA content credentials, detect AI-generated media, and assess manipulation risk. One API call.
Trace C2PA provenance, detect synthetic media, analyze EXIF metadata, and assess manipulation risk.
How it works
Submit a media file or URL and SignalStack runs provenance verification, synthetic media detection, and manipulation analysis in parallel.
1. Submit media
Provide an image or video URL, or upload a file directly. SignalStack extracts frames, metadata, and any embedded provenance manifests for analysis.
2. Multi-layer analysis
The analysis pipeline processes three layers in parallel: C2PA manifest parsing and cryptographic verification, deep learning-based synthetic media classification, and forensic manipulation detection.
3. Provenance report
Receive a comprehensive report with content authenticity score, detection breakdown per model, manipulation heatmaps, C2PA chain visualization, and generator attribution confidence.
from signalstack import SignalStack
client = SignalStack(api_key="sk_...")
result = client.verify.media(
url="https://example.com/photo.jpg",
checks=[
"c2pa_provenance",
"synthetic_detection",
"manipulation_analysis",
],
)
print(f"Authenticity score: {result.score}")
# Authenticity score: 0.15
if result.synthetic:
print(f"AI Generator: {result.synthetic.generator}")
print(f"Confidence: {result.synthetic.confidence}")
# AI Generator: Midjourney v6
# Confidence: 0.97
if result.c2pa:
for action in result.c2pa.provenance:
print(f"{action.actor} → {action.action} at {action.timestamp}")Detection capabilities
Our multi-model ensemble covers synthetic media detection, provenance verification, and manipulation forensics.
C2PA provenance verification
Parse and validate C2PA (Coalition for Content Provenance and Authenticity) manifests. Verify content credentials, hardware-bound signatures, and immutable provenance chains from capture to publication.
Synthetic media detection
Detect AI-generated images and video using deep learning classifiers trained on diffusion models, GANs, and neural rendering pipelines. Identify outputs from DALL-E, Midjourney, Stable Diffusion, Sora, and more.
Manipulation risk scoring
Assess media manipulation risk across multiple dimensions: compression artifacts, splicing detection, resampling analysis, color space anomalies, and histogram inconsistencies.
Deepfake video analysis
Analyze video frames for facial manipulation, lip-sync inconsistencies, temporal artifacts, and generative model fingerprints. Supports MP4, WebM, AVI, and MOV formats up to 4K resolution.
Frequently asked questions
C2PA (Coalition for Content Provenance and Authenticity) is an open standard for cryptographically signing media with provenance metadata. SignalStack parses C2PA manifests, validates the signature chain against hardware-bound certificates, extracts provenance actions (capture, edit, publish), and verifies that the content matches the manifest hash.
Our synthetic media detection uses heuristic analysis of EXIF metadata, C2PA provenance, and file structure to identify potential AI-generated or manipulated content. Specific generator-level attribution is in development.
Our detection combines C2PA manifest verification with forensics analysis of EXIF metadata, MP4 structure, and file-level heuristics. We return a confidence score from 0.0–1.0 based on the combined evidence. Detection accuracy varies by media type and manipulation method.
Yes. We support both file-based analysis and streaming frame analysis via WebSocket connections. Real-time analysis processes up to 30 FPS with sub-second latency per frame for deepfake and manipulation checks.
Yes. For media without C2PA manifests, we perform passive forensic analysis including sensor pattern noise analysis, camera model identification, compression history estimation, and reverse image search to reconstruct provenance where possible.