Detect unsafe or policy-violating content in any image. Supply an image URL (and optional context text) and receive structured flags for adult, violent, hateful, or otherwise restricted content — ideal for pre-screening user uploads before they reach generation pipelines.
About this model
Detect unsafe or policy-violating content in any image. Pass an image URL (and optional context text) and get back boolean flags plus a structured list of categories that were tripped — ideal as a pre-screen on user uploads or AI-generated frames before they enter your downstream pipeline.
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $0.01 per image | Flat fee per moderation call. No volume tiers. |
| Fal.ai | Not available | No equivalent first-party image-moderation model. |
| Replicate | Not available | No equivalent first-party image-moderation model. |
Flat fee per moderation call. No volume tiers.
No equivalent first-party image-moderation model.
No equivalent first-party image-moderation model.
* Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Image URL | string | Image URL to moderate. Public URLs and standard image formats are supported. | https://d3adwkbyhxyrtq.cloudfront.net/ai-images/sample.jpg |
| Context Text | string | Optional context or custom moderation criteria. Use this to bias the moderator with extra context. | |
Image URL to moderate. Public URLs and standard image formats are supported.
https://d3adwkbyhxyrtq.cloudfront.net/ai-images/sample.jpgOptional context or custom moderation criteria. Use this to bias the moderator with extra context.
Developer documentation
Submit the image: POST image_url (a publicly accessible URL) and an optional text to bias the moderator with extra context (for example: "flag anything with weapons").
Poll the result: Call /api/v1/predictions/{request_id}/result until status is completed. Or set a webhook query param to receive the verdict asynchronously.
Act on the verdict: The output.flagged boolean is your gate. output.categories lists which categories were tripped, and output.has_nsfw_contents mirrors the provider's per-output NSFW booleans.
Frequently asked
Any standard web image format reachable by URL — JPEG, PNG, WebP, GIF.
It's an optional context hint for the moderator. Use it to bias the model with custom criteria, for example: "flag content involving children".
Typically completes in a couple of seconds per image.