Enterprise Integration for
Mimir

Photo verification built into Mimir. Right-click any image, get a full forensic verdict — without leaving the CMS.

42 checks  ·  one verdict  ·  plain-language evidence

Monthly Verifications
— remaining
— used — total
Project Status

What's Live

Current state of the Mimir integration. Everything marked "Live" is running in production right now.

https://imagewhisperer.org/axelspringer
Live in production — dedicated enterprise endpoint for Axel Springer editors

Live in Production

Dedicated enterprise endpoint
/axelspringer serves clean editor-focused UI — no pricing, no marketing, no login required
Full analysis pipeline
14 GPU models + Gemini running in parallel. 41 tests per image, ~20s median analysis time
Enterprise usage tracking
Every verification counted against Axel Springer's monthly quota. Admin dashboard at /admin/enterprise
1,500 verifications/month (no paywall)
Enterprise mode bypasses free-tier limits and Stripe payment flow entirely
Priority server access
The server handles up to 8 concurrent analyses. When all slots are busy, public users are queued — enterprise requests from Axel Springer get a guaranteed reserved slot, so editors are never rejected during peak traffic
Known Fakes Database
Instant match in <5ms against previously debunked AI fakes and manipulated images
Fact-check source matching
Cross-references Reuters, Snopes, Politifact, AFP and 20+ fact-check organizations
Web detection & source tracking
Finds where the image has appeared online — exact matches, similar images, partial crops

Integration Details

Native Mimir panel (iframe + postMessage)
ImageWhisperer runs as a tab in Mimir's right panel. Editors right-click a photo, and the image is sent directly for analysis — no manual upload, no context switch.
Two-way postMessage protocol
Mimir sends the image to ImageWhisperer, ImageWhisperer sends the verdict back. Color, label, confidence — ready for Mimir to display or store on the asset.
Quota management
Monthly quota configurable via admin dashboard. Currently set to 1,500/month
CORS & CSP configured for Mimir
frame-ancestors allows *.mimir.tv and *.mjoll.no. CORS headers set for cross-origin postMessage.
REST API (server-to-server)
Available for Mimir REST API to update items with verdicts automatically after analysis completes
Enterprise Exclusives

Built for Mimir

These features are only available when images come through the Mimir integration. They use CMS metadata that public uploads don't have.

Metadata Intelligence

Full CMS EXIF surfacing
Mimir CDN strips EXIF from delivered images. We fetch the original EXIF from Mimir's S3 exifTagsUrl and surface everything: camera, lens, shooting settings, GPS, IPTC, XMP — data not visible in Mimir itself.
Three-badge provenance chain
Every metadata field is tagged with its source: MIMIR (shown in CMS), IMAGE EXIF (from file), or VIA IW (only ImageWhisperer shows it). Editors see exactly which data is new.
Photoshop / editing history detection
Parses XMP History arrays to reveal which software edited the image (Photoshop, Lightroom, GIMP), what actions were taken, and when. Orange warning for Photoshop detections.
7 forensic signals from raw EXIF
Self-declared editing keywords ("bearbeitet", "symbolfoto"), EXIF CompositeImage flag, capture-to-modify time gap, GPS coordinates with map link, usage restrictions, originating DAM system, and artist/photographer data.
Raw CMS metadata panel
Collapsible panel showing ALL fields from ExifTool output, grouped by section (IFD0, ExifIFD, IPTC, XMP, Composite, ICC). Each field badged as "shown above" or "extra" — editors see what's new at a glance.

Editorial Intelligence

Agency trust tier system
Tier 1 (Reuters, AP, AFP, dpa, Getty, EPA) and Tier 2 (NurPhoto, Avalon, Zuma, Sipa, etc.) recognized from CMS credit fields. Trust tier boosts verdict confidence and is displayed in the provenance chain.
EXIF cross-validation
Compares CMS metadata against the image file's own EXIF. Detects "stripped" (CDN removed EXIF), "consistent" (both match), or "divergent" (contradictions found) — an integrity check for image provenance.
CMS name corroboration
When our AI identifies people in the photo, we check IPTC Caption-Abstract, By-line, and Headline from CMS metadata before hedging. If the CMS confirms the name, we say so — no more "unverified AI guess" for subjects documented in wire service captions.
Studio backdrop false-positive guard
Press photos with smooth studio backdrops trigger composite detection (noise/compression mismatch). When CMS confirms it's a press agency photo + B-Free clears, we override the false positive back to green.
Bildunterschrift (auto-caption)
AI-generated wire-ready caption in German, ready for editorial use. Uses the 5W analysis (Wer, Was, Wo, Wann, Warum) enriched by CMS editorial context.
PDF forensic report
One-click download: full provenance chain with source badges, editing history, forensic signals, 5W story, Bildunterschrift, model scores — all in a professional PDF. Designed for editorial archives and legal documentation.
Verdict writeback to Mimir
After analysis, verdict (color, label, confidence) is sent back to Mimir via postMessage. Mimir can store it on the asset or display it in the UI — no manual copy needed.

What Editors See — Provenance Chain

Every metadata field attributed to its source. Data labeled VIA IW is only visible through ImageWhisperer — Mimir's own UI doesn't show it.

Source Agency Reuters MIMIR
Camera Canon EOS R5 IMAGE EXIF
Settings 135mm · f/2.8 · 1/500s · ISO 400 VIA IW
Editing History Adobe Photoshop 25.6 — saved, converted VIA IW
GPS Location 52.5200, 13.4050 → Berlin VIA IW
Integration

How It Works

Right-click a photo in Mimir. ImageWhisperer opens and runs the full analysis automatically. That's it.

1

Right-Click in Mimir

Editor right-clicks any photo in Mimir and selects "Analyze with ImageWhisperer".

2

ImageWhisperer Analyzes

ImageWhisperer opens and automatically starts the full forensic analysis. 14 GPU models + Gemini run in parallel — nothing to upload, nothing to configure.

3

Read the Verdict

Traffic light verdict, confidence score, forensic details, and fact-check sources. The verdict is sent back to Mimir automatically.

Integration Flow

Right-Click Photo
"Analyze with ImageWhisperer"
→
AUTOMATIC
ImageWhisperer Analyzes
14 GPU models + Gemini in parallel
→
Verdict
Traffic light + full forensic report
What happens when the editor right-clicks 
// 1. Editor right-clicks a photo in Mimir
// 2. Selects "Analyze with ImageWhisperer"
// 3. ImageWhisperer opens at:

https://imagewhisperer.org/axelspringer

// 4. The image is sent automatically — analysis starts immediately
// 5. 14 GPU models + Gemini analyze the photo in parallel
// 6. Editor reads the verdict: green (authentic), orange (uncertain), red (AI/manipulated)
// 7. Verification is tracked against Axel Springer's package

Why This Approach Works

One Right-Click
Editor right-clicks any photo in Mimir, selects "Analyze with ImageWhisperer". ImageWhisperer opens and starts the analysis automatically — no upload, no extra steps.
Full Forensic Analysis
14 GPU models + Gemini run 42 independent tests in parallel. AI generation detection, perspective forensics, shadow analysis, metadata forensics, web provenance, fact-check matching, OCR — all in one pass.
Clear Verdict
Traffic light verdict (green/orange/red), confidence score, and detailed forensic report. The editor knows immediately whether the photo is trustworthy.

What We Build — Minimal Effort for Mimir IT

ImageWhisperer delivers the complete verification tool. Mimir's team adds a right-click action that opens ImageWhisperer with the selected photo — that's the entire integration.

We Build (ImageWhisperer)

  • Dedicated /axelspringer panel endpoint (live)
  • postMessage protocol: receive image, return verdict
  • Clean editor UI — no pricing, no marketing
  • Usage tracking per origin for billing & analytics
  • Full forensic analysis (14 GPU models + Gemini)
  • CORS & frame-ancestors configured for *.mimir.tv

Mimir IT Has Done

  • Right-click action: "Analyze with ImageWhisperer"
  • Action opens ImageWhisperer with the selected photo
Integration complete — live in production
Infrastructure

Dedicated Hardware

ImageWhisperer runs on dedicated server hardware in a German datacenter. No shared cloud instances — your data stays in the EU.

Processor
Intel Core i5-13500
13th Gen, 20 threads. Runs all 14 detection models in parallel without bottleneck. Enough headroom for 1,400+ analyses per hour.
GPU
NVIDIA RTX 4000 SFF Ada
20 GB VRAM. Professional workstation GPU that accelerates the deep learning models (B-Free, SPAI, TruFor, IML-ViT, etc.) for sub-second inference.
Memory
64 GB DDR5 RAM
All 14 GPU models stay loaded in memory — zero cold-start latency. No model swapping, no waiting. Every analysis starts immediately.
Storage
2× 1.7 TB Samsung NVMe
Enterprise-grade Samsung SSDs in RAID. Fast image I/O, analysis cache, and Known Fakes Database with sub-millisecond lookups.
Dedicated datacenter, Falkenstein, Germany · GDPR-compliant · Daily encrypted backups · No data leaves the EU
Pricing

Chosen Package

Axel Springer selected the Editorial Quick Check package for the pilot. Upgrade paths available as usage grows.

Active Pilot

Editorial Quick Check

~50 checks/day · ~1,500/month
€500
per month · €0.33 per verification
  • Full pipeline: 14 GPU models + Gemini AI
  • Dedicated /axelspringer endpoint
  • Editor triggers check manually
  • Known Fakes Database (<5ms instant match)
  • EXIF & metadata forensics
  • Error Level Analysis (ELA)
  • Composite image detection
  • Mimir integration & ongoing updates included
Partner Deal
All features of the highest tier (Photo Desk Standard) are activated — including location verification, shadow analysis, perspective forensics, C2PA detection, claim verification, and logo/landmark recognition.
Response: 10–30 seconds
Upgrade Option

Newsroom

~130 checks/day · ~4,000/month
€625
per month · €0.16 per verification
  • Everything in Quick Check, plus:
  • 2.7× the volume (4,000/month)
  • Web detection & reverse image search
  • Fact-check cross-referencing (20+ organizations)
  • OCR text extraction (6 languages)
  • QR code & barcode extraction
  • Short URL resolution (bit.ly, t.co, goo.gl, etc.)
  • AI generator identification (Midjourney, DALL-E, Flux, etc.)
Response: 10–30 seconds
Upgrade Option

Photo Desk Standard

~250 checks/day · ~7,500/month
€750
per month · €0.10 per verification
  • Everything in Newsroom, plus:
  • 5× the volume (7,500/month)
  • Location verification & GPS cross-check
  • Shadow & lighting consistency analysis
  • Vanishing point & perspective forensics
  • Text-on-signs manipulation detection
  • C2PA / Content Credentials detection
  • Claim verification via Google Custom Search
  • Logo & landmark recognition
Response: 10–30 seconds · 41 tests per image

What Happens Beyond the Package Limit?

We never block editors. If a newsroom hits its monthly limit, verifications keep working — no interruption, no lockout.

Soft Limit — Overage Rate
Verifications beyond the package are billed at +20% of the normal per-check price for that package. Settled at the end of the month.
Example: Editorial Quick Check includes 1,500 checks at €0.33.
Check #1,501 onwards: €0.40 per verification (0.33 × 1.20).
Upgrading — On Request
Want a higher package? Request an upgrade and it takes effect at the start of the next billing month. The current month finishes on the existing package, with any overage at the +20% rate.
No automatic upgrades — you decide when to move up.
No surprises. We send a notification when usage reaches 80% and 100% of the package limit, so editors and managers can plan ahead.
Both parties can cancel with 90 days notice. No long-term lock-in.

Volume Estimates from Eva Bauer

Scenario Per Day Per Month Cost/Check Overage Rate Package
Manual check by editor (minimum) 50 ~1,500 €0.33 €0.40 Quick Check
Mixed editorial + breaking news 130 ~4,000 €0.16 €0.19 Newsroom
All manual uploads, automated (maximum) 250 ~7,500 €0.10 €0.12 Photo Desk
All agency + manual (not recommended) 40,000 ~1.2M — — Not practical

Why not a simple AI score?

A single AI-detection API gives you one number. That works — until it doesn't. Every model has blind spots: one misses Flux-generated images, another flags real photos with bokeh as "manipulated," a third can't tell a photoshopped composite from a clean original.

ImageWhisperer runs 42 independent tests on every image — neural forensics, metadata analysis, web provenance, fact-check lookup, camera fingerprinting, and more — then cross-validates them against each other. No single model can override the verdict alone; every flag requires corroboration. That's what turns a score into a decision your editors can trust under deadline pressure.

21
Independent tests per image
10+
GPU models cross-validated
0
Single-model overrides allowed

Building this in-house means licensing multiple detection APIs, running GPU servers, writing the orchestration and cross-validation logic, and keeping up as new generators emerge monthly. Our per-image price reflects that full pipeline — not a single API call.

Pilot Active

The pilot is live.

The integration is running in production. Questions, feedback, or ready to scale up? Reach out anytime.

Get in Touch