AI-generated Answer for noah.jackson
Short answer: there’s no officially documented profanity filter in Scribe v2 Realtime. The WebSocket API reference does not list any parameter or behavior that censors or masks profanity; it streams partial and committed transcripts verbatim via events like partial_transcript and committed_transcript. (elevenlabs.io)
What to check:
- Reproduce with the bare Realtime WebSocket endpoint (model_id=scribe_v2_realtime) and inspect the committed_transcript text. If curse words appear there, any masking is happening in your app layer (UI, logging, or a downstream component), not in Scribe. (elevenlabs.io)
- If you’re testing through another ElevenLabs surface, note that “Live Moderation” applies to some shared voices in the Voice Library (TTS), not to Speech-to-Text. That feature can block certain textual inputs but isn’t part of Scribe Realtime. (help.elevenlabs.io)
- For product context, Scribe v2 Realtime is the low‑latency live STT model; the docs and help center pages describing Speech‑to‑Text do not mention profanity filtering. (elevenlabs.io)
If you can share the session_id from the session_started event (returned by the Realtime API), support can investigate a specific run. (elevenlabs.io)
Sources:
Realtime | ElevenLabs Documentation
Realtime | ElevenLabs Documentation
What is Live Moderation? – ElevenLabs
Transcription | ElevenLabs Documentation
Realtime | ElevenLabs Documentation
feral hearth
barren creek