Content Moderation
Bria's APIs are built based on Bria's proprietary foundation models trained exclusively on licensed, safe-for-commercial-use data. By training on carefully curated datasets that exclude harmful, copyrighted, and privacy-infringing content, our models are explicitly designed to avoid generating such content. As an additional layer of protection, we've introduced content moderation capabilities that can be enabled on demand across our API endpoints.
Overview
Content moderation is available as an opt-in feature, controlled by the content_moderation parameter (default: false). When enabled, it provides automated screening of both input and output visuals across all supported endpoints.
Moderated Categories
Our moderation system screens for the following categories:
- Explicit Content: Explicit nudity, exposed genitalia, buttocks, and female nipples; explicit sexual activity; and sex toys
- Non-Explicit Nudity: Bare backs, partially exposed intimate parts, implied nudity, obstructed intimate parts, and kissing on the lips
- Swimwear/Underwear: Male and female swimwear or underwear
- Violence: Weapons, weapon violence, physical violence, self-harm, blood and gore, and explosions and blasts
- Visually Disturbing Content: Emaciated bodies, corpses, and crashes (e.g., air crashes)
- Substances: Depictions of pills, smoking, alcohol use, and alcoholic beverages
- Offensive Gestures: Middle finger gestures
- Gambling: Casino games and gambling activities
- Hate Symbols: Nazi imagery, white supremacy symbols, extremist flags
Implementation Details
Input Validation
- Any input image failing moderation triggers immediate request termination
- Returns a 422 error with details about the failed parameter
Output Handling
Behavior varies based on request type:
Synchronous Requests (sync=true)
- Single output: Returns 422 error if moderation fails
- Multiple outputs: Returns successful generations while indicating blocked content
Asynchronous Requests (sync=false)
- Blocked content is replaced with zero-byte files at the placeholder URLs
- Valid content remains accessible at the provided URLs
API Response Examples
Our API maintains consistent response patterns across endpoints while accommodating content moderation results. Check individual endpoint documentation for specific response formats.
For more detailed information about specific endpoints and their implementation of content moderation, refer to the relevant API documentation sections below.