~/docs cat which-input.md
Which input do I send?
The server infers the media type from the bytes. You never name a technique. Pick the row that matches what you are holding.
| Input | Send | Formats | Size cap | Credits | Evidence in the reply |
|---|---|---|---|---|---|
| Image | file | JPEG, PNG, WEBP, GIF, BMP | image ≤ 10 MB | 1 credit | None. The verdict and score cover the whole image. |
| Video | file or url | MP4, MOV, WEBM, MKV, AVI | video ≤ 100 MB | 1 per scored frame, at most 20 | frames[]: one entry per sampled frame with its score, plus threshold_used. |
| Audio | file | MP3, WAV, FLAC, M4A, AAC, OGG | audio ≤ 50 MB | 1 per minute, rounded up, minimum 1 | segments[]: one value per window, plus duration_ms. |
| Link | url | A public video page, or a direct link to a video file | video ≤ 100 MB once fetched | The same as a video: 1 per scored frame, at most 20 | source: where it came from and how much was scored, plus the video's frames[]. |
| Document | file | PDF; .doc, .docx and .tif are also accepted by extension | document ≤ 50 MB | Needs a document plan. Without one the call answers 402. | Contact sales to enable document detection. |
A file over its cap answers 413. A file no detector can read answers 422 with UNSUPPORTED_MEDIA_TYPE. Costs, caps and rate limits are collected on Credits and limits.