An operation is an atomic action performed by our engine. Operations are used to track usage levels across time and across products. Each API call that consumes operations will return a specific field to let you know how many operations were used.
The dashboard lets you to track your usage levels, including the operations consumed.
Visual Moderation
For Visual Moderation, the number of operations consumed will depend on the models chosen.
Models that belong to the same group count as a single operation. For instance, using the recreational_drug, alcohol and tobacco models in a given request will count as 1 operation:
| Group | Models | Operations |
|---|---|---|
| All nudity models | nudity nudity-1.0 nudity-1.0 nudity-2.0 nudity-2.1 | 1 |
| Violence, Gore, Weapon, Self-harm | weapon gore gore-2.0 violence self-harm | 1 |
| Face / People models | faces face-analysis | 1 |
| Face / People models | face-age | 1 |
| Restricted content | recreational_drug alcohol tobacco medical gambling money | 1 |
| Hate & Offensive | offensive offensive-2.0 | 1 |
| AI-generated media detection | genai | 5 |
| Deepfake detection | deepfake | 5 |
| OCR, Text, QR codes | text text-content qr-content ocr | 1 |
| Sensitive scenes | destruction military | 1 |
| Content description | type properties | 1 |
| Quality assessment | quality | 1 |
Please note that this pricing may not apply to accounts on legacy models (from before 2025).
For Videos, GIF images and batch requests, the number of operations is multiplied by the number of frames analyzed.
Video Moderation defaults at 1 frame analyzed every 2 seconds, but you can change this setting.
Text Moderation
For Text Moderation, 1 operation is consumed when you submit a text item containing fewer than 1000 characters. An extra operation is consumed for each extra 1000 characters. So if you do a request with a text that is 1800 characters long, that will count as 2 operations.
Was this page helpful?