ElevenLabs releases Scribe v3 speech to text and gives sound effects their own concurrency limits (November 9, 2026)

What changed

ElevenLabs posted its changelog on November 9, 2026. The parts developers need to act on:

• Scribe v3. The Speech to Text API accepts scribe_v3 as model_id for recorded audio, and the realtime WebSocket accepts scribe_v3_live. ElevenLabs says Scribe v3 makes 27% fewer word errors than Scribe v2 on the FLEURS benchmark. Scribe v2 and Scribe v2 Realtime remain available.
• New Scribe v3 parameters: numbers_format (digits or words) and secondary_language_codes for audio that mixes languages. Scribe v3 Live adds commit_strategy=turn_prediction, multilang_mode and a keyterm library of up to 10,000 terms.
• Sound effects concurrency. Sound effects no longer share the Text to Speech limit. The limits are 4 on Free and Starter, 6 on Creator and Pro, and 8 on Scale and Business. Requests beyond the limit and its reduced-priority queue return 429.
• Realtime API audio formats. audio/pcma is removed from the audio format list, and audio/pcm input at sample rates other than 24000 Hz is rejected with invalid_audio_format.
• Conversation history. Requests from users who can see more than 10,000 agents return 403 conversation_history_agent_limit_exceeded unless they filter by agent.
• Credit limits. PUT /v1/workspace/members/{user_id}/usage-limit sets a member's or service account's monthly credit limit (usage_limit), or removes it with null.
• Gemini fallback. The fallback ramp for agents on gemini-2.5-flash and gemini-2.5-flash-lite now starts November 15 instead of November 8. The November 30 cutoff is unchanged.
• SDKs. JavaScript and Python SDK v2.77.0 add a Workflow builder for agent workflows in code.

Prerequisites

• An ElevenLabs API key in ELEVENLABS_API_KEY