Detect AI-Generated Images, Audio, and Text with Open-Source Models

A new tutorial by Emeka Okafor demonstrates building a local Python pipeline that uses three open-source Hugging Face classifiers to detect AI-generated images, audio, and text, emitting a JSON verdict and exit code to gate uploads. The models include Organika/sdxl-detector for images (98% validation accuracy, CC-BY-NC-3.0), MelodyMachine/Deepfake-audio-detection-V2 for audio (99.7% accuracy, Apache-2.0), and fakespot-ai/roberta-base-ai-text-detection-v1 for text (Apache-2.0). The pipeline runs on CPU, requires Python 3.10+, transformers 5.15.0, PyTorch 2.13.0, and about 2 GB disk space, with a caution that detector scores are risk signals, not ground truth.

Detect AI-Generated Images, Audio, and Text with Open-Source Models Wire three Hugging Face classifiers into one local pipeline that flags synthetic media before it reaches users. Emeka Okafor https://sourcefeed.dev/u/emeka okafor What you'll build A local Python pipeline that routes incoming files to three open-source classifiers — one each for images, audio, and text — and emits a JSON verdict synthetic or authentic plus a non-zero exit code you can use to gate uploads before they reach your users. Prerequisites - Python 3.10+ verified with 3.12; Transformers https://huggingface.co/docs/transformers v5 requires ≥3.10 transformers 5.15.0, PyTorch https://pytorch.org 2.13.0 Transformers v5 needs torch ≥2.5 , Pillow FFmpeg https://ffmpeg.org on your PATH — the audio pipeline shells out to it to decode files- ~2 GB free disk: the three model checkpoints total roughly 1.2 GB on first download - No GPU required. Everything here runs on CPU; a GPU just makes it faster. - macOS or Linux assumed for shell commands; everything works on Windows with the usual venv path changes 1. Set up the environment mkdir synthetic-gate && cd synthetic-gate python3 -m venv .venv && source .venv/bin/activate pip install "transformers==5.15.0" torch pillow Install FFmpeg if you don't have it: macOS brew install ffmpeg Debian/Ubuntu sudo apt-get install -y ffmpeg 2. Meet the three classifiers You want one battle-tested checkpoint per modality, all loadable through the same pipeline API so the routing code stays trivial: Images: — a Swin transformer fine-tuned on Wikimedia-vs-SDXL image pairs, ~98% validation accuracy. Labels: Organika/sdxl-detector artificial / human . Heads up: it's licensed CC-BY-NC-3.0, so it's fine for evaluation and internal research but not commercial deployment — for that, swap in the Apache-2.0 EfficientNet-B4, loads via Dafilab/ai-image-detector timm instead of pipeline . Audio: — wav2vec2-base fine-tuned for spoofed-voice detection, 99.7% accuracy on its eval set, Apache-2.0. Labels: MelodyMachine/Deepfake-audio-detection-V2 fake / real . Expects 16 kHz input; the pipeline resamples for you when FFmpeg decodes the file. Text: — a RoBERTa-base classifier from Fakespot Mozilla for English text, Apache-2.0. Labels: fakespot-ai/roberta-base-ai-text-detection-v1 Human / AI . RoBERTa caps at 512 tokens, so longer inputs must be truncated or chunked. One thing to internalize before shipping any of this: detector scores are risk signals, not ground truth. All three models were trained against specific generator families, and accuracy drops on generators released after training, on heavily compressed re-uploads, and on out-of-domain content. Route high scores to a human review queue; don't auto-ban on them. 3. Write the routing pipeline Save this as detect.py : python import json import sys from pathlib import Path from transformers import pipeline MODELS = { "image": "image-classification", "Organika/sdxl-detector" , "audio": "audio-classification", "MelodyMachine/Deepfake-audio-detection-V2" , "text": "text-classification", "fakespot-ai/roberta-base-ai-text-detection-v1" , } EXTENSIONS = { ".jpg": "image", ".jpeg": "image", ".png": "image", ".webp": "image", ".wav": "audio", ".mp3": "audio", ".flac": "audio", ".m4a": "audio", ".txt": "text", ".md": "text", } Each model names its labels differently; anything in this set means "machine-made". SYNTHETIC = {"artificial", "fake", "AI"} pipes = {} def get pipe modality : if modality not in pipes: task, model = MODELS modality pipes modality = pipeline task, model=model return pipes modality def detect path: Path - dict: modality = EXTENSIONS.get path.suffix.lower if modality is None: raise SystemExit f"unsupported extension: {path.suffix}" pipe = get pipe modality if modality == "text": truncation is forwarded to the tokenizer; RoBERTa can't take 512 tokens scores = pipe path.read text encoding="utf-8" , truncation=True, top k=None else: scores = pipe str path best = max scores, key=lambda s: s "score" return { "file": path.name, "modality": modality, "verdict": "synthetic" if best "label" in SYNTHETIC else "authentic", "label": best "label" , "score": round best "score" , 4 , "model": MODELS modality 1 , } def main : results = detect Path p for p in sys.argv 1: print json.dumps results, indent=2 exit 1 if anything is confidently synthetic, so CI or an upload hook can gate on it if any r "verdict" == "synthetic" and r "score" = 0.90 for r in results : sys.exit 1 if name == " main ": main The pipelines are cached in pipes so each model loads once per process — model load dominates runtime, so in production you'd keep this warm behind a small HTTP service rather than spawning a process per file. The 0.90 gate threshold is a starting point; tune it against your own false-positive tolerance. 4. Feed it real inputs Collect one known-real and one known-synthetic sample per modality: a photo off your phone plus an image from any diffusion model; a voice memo plus a clip from a TTS service; a paragraph you wrote plus one straight out of a chatbot pasted into llm.txt . python detect.py phone-photo.jpg diffusion-render.png voicememo.wav llm.txt First run downloads all three checkpoints from the Hugging Face Hub ~1.2 GB into ~/.cache/huggingface ; later runs load from cache and classify each file in a few hundred milliseconds to a few seconds on CPU. Verify it works You should see one JSON object per file, with the synthetic samples flagged: { "file": "phone-photo.jpg", "modality": "image", "verdict": "authentic", "label": "human", "score": 0.9971, "model": "Organika/sdxl-detector" }, { "file": "diffusion-render.png", "modality": "image", "verdict": "synthetic", "label": "artificial", "score": 0.9998, "model": "Organika/sdxl-detector" }, { "file": "voicememo.wav", "modality": "audio", "verdict": "authentic", "label": "real", "score": 0.9842, "model": "MelodyMachine/Deepfake-audio-detection-V2" }, { "file": "llm.txt", "modality": "text", "verdict": "synthetic", "label": "AI", "score": 0.9564, "model": "fakespot-ai/roberta-base-ai-text-detection-v1" } Your exact scores will differ. Then confirm the gate fired: echo $? → 1, because at least one file scored ≥0.90 synthetic Run it again with only the real samples and echo $? should print 0 . Troubleshooting ValueError: ffmpeg was not found but is required to load audio files from filename — the audio pipeline decodes files by invoking FFmpeg. Install it brew install ffmpeg / apt-get install -y ffmpeg and make sure it's on the PATH of the process running the script, which is easy to miss inside slim Docker images and systemd units. Token indices sequence length is longer than the specified maximum sequence length for this model 743 512 followed by IndexError: index out of range in self — you dropped truncation=True from the text call. Restore it; and remember truncation means only the first 512 tokens are scored, so chunk long documents and score each chunk if you care about the tail. OSError: We couldn't connect to 'https://huggingface.co' to load this file — first run needs Hub access. If your servers are egress-restricted, pre-download on a machine that isn't hf download Organika/sdxl-detector , once per model — the hf CLI ships with huggingface hub , copy ~/.cache/huggingface across, and set HF HUB OFFLINE=1 in production. Everything scores authentic, even known fakes — check the input actually reached the model intact. Screenshots of AI images, re-encoded thumbnails, and voice notes transcoded by a chat app strip the artifacts detectors key on. Test with the original file, and treat detector evasion via re-encoding as a known limitation, not a bug in your wiring. Next steps Single-model verdicts are the weakest form of this pipeline. To harden it: ensemble two or more detectors per modality and require agreement before flagging; add provenance checks with C2PA Content Credentials https://c2pa.org so signed-authentic media can skip ML scoring entirely; and benchmark on public datasets CIFAKE and GenImage for images, ASVspoof for audio before trusting any threshold. When you outgrow the CLI, wrap get pipe / detect in a FastAPI service with the pipelines loaded at startup, and log every score — your own traffic distribution is the fine-tuning dataset that will eventually beat any off-the-shelf checkpoint. Sources & further reading - Organika/sdxl-detector model card https://huggingface.co/Organika/sdxl-detector — huggingface.co - MelodyMachine/Deepfake-audio-detection-V2 model card https://huggingface.co/MelodyMachine/Deepfake-audio-detection-V2 — huggingface.co - fakespot-ai/roberta-base-ai-text-detection-v1 model card https://huggingface.co/fakespot-ai/roberta-base-ai-text-detection-v1 — huggingface.co - Transformers Pipelines documentation https://huggingface.co/docs/transformers/main classes/pipelines — huggingface.co - Hugging Face hf CLI guide https://huggingface.co/docs/huggingface hub/guides/cli — huggingface.co - transformers on PyPI https://pypi.org/project/transformers/ — pypi.org Emeka Okafor https://sourcefeed.dev/u/emeka okafor · Security Editor Emeka has spent over a decade tracking threat actors, vulnerability disclosures, and the evolving landscape of application security, bringing a sharp continent-spanning perspective to his reporting. He's known for translating dense CVE advisories into clear, actionable context that developers and security teams alike actually read. Discussion 0 No comments yet Be the first to weigh in.