Contact info on product images
Sellers embedding phone numbers and off-platform links in listings
Phone number detected
Detect text in images with OCR moderation
Moderate text in images. Lasso's OCR moderation detects contact information, watermarks, and circumvention by hidden text in images.
Evasion links in profile pictures
Hidden URLs in gaming avatars used to circumvent platform rules
Evasion URL detectedTelegram handles in profile photos
Contact info embedded in selfies to move users off-platform
Contact info detectedSlurs in meme text
Obfuscated hate speech like H4TE SPE4K in user-uploaded images
Text extractedLeet speak flagged
OCR moderation: four layers to catch what text filters miss
1OCR extraction
Fast, multilingual extraction at high quality.
Lasso's OCR detects any kind of text in images.
See Lasso in action
Text region detected
OCR Extraction
3 regions found
Line 1
txt me directSolicitation
Line 2
555-0192Phone
Line 3
better priceText
Extraction confidence97%
2Custom rules
Use extracted text from OCR in our proprietary rule engine, configured to your platform standards.
Filter out anything your specific platform rules do not allow.
See Lasso in actionLayer 1 Output
1txt me direct
2555-0192
3better price
Custom Rules
BlockPersonal information
Phone numbers
Email addresses
Home addresses
Social media handles
Blocked
Phone number detected in extracted text
3AI moderator
AI Moderator understands the context of images.
Catches policy violations in meme captions and overlays. Routes to auto-remove, flag for review, or approve.
See Lasso in actionEmbedded text detected
AI Moderator
Analyzing
Extracted text
l1nktr.ee/DM4more
Decoded
linktr.ee/DM4more
Pattern recognized
Linktree URL off-platform redirect
Classification
Evasion URL hidden in avatar artwork
Auto-removed
Evasion URL in profile image artwork
4Human review
Gray area content queues for human review.
Each item shows: extracted text, original image, AI reasoning, and suggested action. Every human decision trains the system.
See Lasso in actionQueued for review
AI Moderator reasoning
Possible evasion URL: l1nktr.ee/DM4more — character substitution pattern detected.
Human Review1 of 3
Extracted text
l1nktr.ee/DM4more
AI assessment
Likely evasion URL — Linktree redirect via character substitution
AI confidence
Evasion URL74%
Below auto-action threshold (90%)
This decision will train the AI for future similar content
Three capabilities you get with OCR moderation.
Lasso detects 99.9% of text in images.
@username
Type a message…
Extracted handles
@username
discord ID
email
Detection rate
99.9%
200+ languages including CJK, Arabic, Cyrillic, from one unified model.
People move fast, so does our OCR.
@username
Type a message…
Extracted handles
@username
discord ID
email
Detection rate
99.9%
Lasso: next-gen AI content moderation
99%
On autopilot, and getting smarter every day.
Three layers of AI handle the volume, your rules, and the grey areas. Your team only sees what truly needs them, and every decision they make improves the system.
Keep users safe without driving them away.
Customizable moderation that lets you find the right balance between safety and user experience. So you protect your community without suppressing the culture that makes it worth joining.
Complexity removed from content moderation.
One API. Clear dashboards. A moderation pipeline built around one-click actions and the right context, right where you need it.
★★★★★
4.9
Highest rated in content moderation on G2.
Every industry has its evasion tactics. OCR moderation finds them.
Gaming
Players embed obfuscated slurs in memes and screenshots to evade chat filters.
- In-game screenshots with hidden slurs
- Profile banners with obfuscated text
- Memes with leet speak insults
Dating
Romance scams start with off-platform contact.
- Phone numbers hidden in profile photos
- Social handles embedded in images
- Payment addresses in profile pics
Marketplaces
Sellers embed URLs, payment handles, and contact info in product images.
- Hidden URLs in product images
- Venmo/payment handles in photos
- External marketplace links
FAQs
OCR moderation, answered
99% detection accuracy on policy violations. 99.9% text extraction accuracy across 200+ languages.
Yes. Confidence scores range 0–100. Set your automation threshold per policy: high-confidence flags auto-action, borderline flags route to human review. You tune it per policy and per industry.
Lasso's OCR extracts text in 200+ languages and scripts. CJK, Arabic, and Devanagari are individually optimized. You can configure different moderation rules per language if needed.
Yes. Common techniques detected: leet speak (c0d3 for code), homoglyph substitution (Cyrillic vs Latin lookalikes), strikethrough formatting, tiny text, QR codes redirecting to external sites, payment addresses in images. Your custom models define what counts as evasion.
Text extraction in realtime, no delay experienced by users. At millions of images, Lasso maintains predictable latency for real-time moderation.
See what's hiding in your images
Lasso's OCR moderation catches the text that bypasses every other filter. Set up in minutes, not months.
Book a demo
Lasso Moderation
© 2026. All rights reserved.
