Run a photo or video through the lab.
Phenomainon runs a battery of independent pipelines on your upload: AI object detection, a custom aircraft classifier trained on our own datasets, mundane-object findings, scene context, a computed sky report (planets, bright stars, satellite passes at the capture place and time), cloud identification, a lens-artifact check, motion-event tracking, and forensic authenticity checks against metadata, frame duplication, scene cuts, and audio splices.
Every run also includes Multimodal AI Analysis — a frame-by-frame AI review alongside our own pipelines. Nothing is stored — the file is deleted as soon as we read it, unless you choose to save it with an identification suggestion on an Unknown result.
Drag everything from the incident here — or hit Choose files and multi-select it all: video, photos, even the phone's data files. We sort it automatically: the video is analyzed, data files become phone data, and the rest ride into the filed case. Up to 500 MB each.
Have data from the phone? Include it in Choose files with everything else — it routes to the phone-data slot automatically — or attach it here separately. One file, several, or a zip: a Google Takeout export, an iPhone “export unmodified originals” folder (media + .AAE), a GPS/sensor log (GPX, DJI SRT, CSV/JSON), XMP sidecars. Everything inside is scanned and checked against what the media file says about where, when, and on what it was recorded — and a bundled copy of the clip is byte-compared to your upload.
Forensics (optional) reads the file's structure — container, capture keys, encoder fingerprints, timeline — and answers “camera-original, exported, or stitched?” with every piece of evidence, plus one combined authenticity score. Works for photos and videos.
Every analysis includes Multimodal AI Analysis — one frame is reviewed by a frontier multimodal model alongside our own pipelines.
- Up to 500 MB. Files under 4 MB stream straight to the proxy. Larger files upload to blob storage and run through an async job queue with progress polling — same result, just longer wait.
- Videos take time. Inference runs on CPU. A 30–90 second video can take tens of seconds.
- Authenticity is not deepfake detection. It flags editing and handling artifacts — signs of cuts, re-encoding, missing provenance — not AI-generated content.
- “Unknown” means unknown to this model. The classifier knows a fixed set of aircraft types. When margin, spread, and frame agreement don’t support a match it reports Unknown and shows the closest shape as context only — that is a statement about the model’s limits, not a claim about what the object is.
- Similar airframes can read alike. Measured on held-out training images, lookalike types still get confused — a Rafale can read as a JAS-39, a B-21 as a B-2. Corroborate anything that matters.