fakespot-ai-roberta-base-ai-text-detection-v1-browser

RoBERTa AI Text Detector — Browser ONNX (Lynote mirror)

Browser-ready ONNX export of the fakespot-ai/roberta-base-ai-text-detection-v1 AI text detector (125M, RoBERTa-base), converted by Lynote for 100% in-browser inference with transformers.js (WASM/WebGPU, no API key, no server).

Attribution

• Upstream model: fakespot-ai/roberta-base-ai-text-detection-v1 (Apache-2.0)
• Base model: FacebookAI/roberta-base (Apache-2.0)
• Conversion: exported to ONNX (opset 14, fp32) by Lynote from the StealthHumanizer project (scripts/model/export-browser-models.py). Lynote does not claim authorship of the upstream weights.

This repo mirrors the upstream model in ONNX format for deployment convenience; the model card, training details and licenses of the original repos remain authoritative.

Usage (transformers.js)

import { pipeline } from '@huggingface/transformers';
const detector = await pipeline('text-classification', 'Lynote/fakespot-ai-roberta-base-ai-text-detection-v1-browser');
const out = await detector('This text looks like it was written by a machine.');

Or via the StealthHumanizer app with NEXT_PUBLIC_BROWSER_MODEL_BASE=https://huggingface.co/Lynote.

Interpret cautiously

Detector output is a probabilistic signal, not proof of authorship. Do not use it alone for disciplinary, legal, or moderation decisions.