A unified API for detecting AI-generated text and images. Multi-stage ML pipeline with Binoculars, RoBERTa, and ViT classifiers. Built for developers, newsrooms, and compliance teams.
100 scans/month free · No credit card required
97%+
Text detection accuracy
<600ms
Median latency (GPU)
20+
Languages supported
3-stage
Detection pipeline
Send text or images to a single REST endpoint. The pipeline automatically selects the right detectors and returns a structured result with per-stage scores and a plain-English explanation.
Request
curl -X POST https://truthlens-api.fly.dev/v1/detect \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"modality": "text",
"input_type": "raw_text",
"content": "The quick brown fox jumps over the lazy dog."
}'Response
{
"id": "a1b2c3d4-...",
"verdict": "human",
"confidence": 0.12,
"explanation": "Content shows human-like writing patterns.",
"details": {
"heuristic_score": 0.18,
"binoculars_score": 0.09,
"roberta_score": 0.11,
"stages": {
"heuristics": true,
"binoculars": true,
"roberta": true
}
},
"latency_ms": 547
}A full detection stack — not just a classifier.
Multi-stage pipeline: heuristics + Binoculars (Falcon-7B) + RoBERTa. Works on 20+ languages.
Detect AI-generated images from DALL-E, Midjourney, Stable Diffusion via metadata, FFT, and ViT classifier.
Process up to 100 items per request. 10 concurrent workers. Results returned in input order.
Automatic language detection. Supports English, Spanish, French, Arabic, Chinese, Japanese, and more.
API key management, rate limiting, usage analytics, and per-request scan history.
Every result comes with per-stage scores, feature breakdowns, and a human-readable explanation.
Start free. Pay only when you grow.