Decart "moderation_violation" and Ended Sessions: Why Lucy Refuses Requests and What to Do

When a Decart session ends with "moderation_violation", the request was refused by a content filter. What that means, what usually triggers it, and practical next steps.

What the error means

Decart's realtime SDK ends a session with a reason code. The ones people search for:

• moderation_violation or policy_violation: the content filter refused the input or prompt, and the session closes. The SDK deliberately doesn't retry, so the same content isn't sent again.
• insufficient_credits: the account ran out of credit.
• session_limit: too many sessions at once for the account.

Only the first two are refusals. The others are billing and capacity limits.

What commonly triggers AI video refusals

Every AI video provider runs filters, and they tend to be cautious around:

• Recognisable public figures or celebrities as the target face
• Prompts describing violence, nudity or minors
• Reference photos that look like they were taken from someone else's profile
• Text in the prompt that reads like impersonation ("make me look exactly like…")

Filters err on the side of refusing, so honest requests sometimes get caught too.

What to do

1. Rephrase neutrally. Describe a style or look instead of naming a real person.
2. Use your own face, or one you have permission to use, as the reference. That's also the rule that keeps you legally safe (is face swap legal).
3. Use a clear, front-facing photo without filters or heavy edits (the reference photo guide).
4. Check the error type. "insufficient_credits" and "session_limit" aren't refusals: top up or close other sessions.

A simpler option for everyday face swap

If you just want to look like a different (consenting) person on camera, or swap faces in your own clip, you don't need a developer API at all. the FaceLive studio runs in the browser with ready-made characters, and the FaceLive Camera for Windows app puts the result straight into WhatsApp, Zoom or Teams. FaceLive also refuses impersonation and abuse, but you won't need to engineer prompts for normal use.

FAQ

Why did Decart refuse my face swap? Its content filter flagged the input or prompt. Celebrity faces, explicit prompts and impersonation wording are common triggers.

Will retrying help? Not with the same content. Change the reference or wording.

Is there a face swap that never refuses? No legitimate one. Every responsible service blocks impersonation and abuse. Using your own face or one you have permission to use avoids almost all refusals.