WHAT'S NEW
Public log of what's shipped recently and where this is heading. New entries appear at the top.
Audio detection arrives
Drop in an MP3, WAV, FLAC, M4A, OGG, or Opus clip and it gets scored the same way videos and images already are. The detector now looks for tell-tale ID3 / iTunes tags, the sample rate / channel layout fingerprint typical of AI text-to-speech, and direct signatures from ElevenLabs, Suno, Udio, Bark, OpenAI TTS, and a dozen others.
Image detection arrives
Photos and PNG/WebP/HEIC images can now be analyzed. EXIF make/model, ICC profile presence, XMP creator-tool tags, JPEG quantization-table fingerprint, and direct Midjourney / Stable Diffusion / DALL-E / Adobe Firefly / Sora signatures all feed the rule engine. Real camera photos with their full metadata intact stay easy to distinguish from AI renders with sparse or telltale metadata.
New home on ironcatlabs.com
DFD moved off its private subdomain to its own public home at dfd.ironcatlabs.com. Same detector, same corpus, fresh certificate, and a cleaner separation between this project and the internal tools it grew alongside.
Rules now snapshot per upload
The detector's rule set used to recompute live on every analysis. It now snapshots the rule state every time a verified sample is added to the training corpus, so verdicts are reproducible and the history of how the model evolves is visible. Every analysis records which rule snapshot scored it.