ElevenLabs AI Updates: November 5, 2026
Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled
1. SFX v3, Eleven Realtime and Eleven v4 Voice Clone Fine-Tuning Reach Every ElevenLabs Workspace
ElevenLabs posted its changelog for the week of November 2, which opens SFX v3 (eleven_text_to_sound_v3), previously in limited access, to every workspace and makes it the default model in Sound Effects, Studio and Flows, as well as for the sound effects that Video to music adds; API callers select it with model_id on Create sound effect and can set loop to true, and eleven_text_to_sound_v2 is now deprecated and will be removed on March 1, 2027. Eleven Realtime, whose gradual rollout began on October 19, is now available to ElevenAgents voice agents in every workspace: an agent can use Eleven Realtime v1 Mini or v1 Max as its voice mode, or an OpenAI Realtime client can connect to the Realtime WebSocket, and realtime conversations can be grouped through /v1/convai/realtime/conversation-groups. Idle realtime audio sessions now close after 3 minutes instead of 60 seconds, with a session_idle_timeout error. Professional Voice Clone fine-tuning for Eleven v4 and Eleven v4 Turbo is also open to every workspace, and requests to a v4-tuned clone now use the clone_strength saved in the voice’s settings when a request sets neither clone_strength nor use_pvc_as_ivc. Source
2. ElevenAgents Adds GPT-6.1 Terra, Defaults New Agents to Gemini 3.5 Flash, and Sets GPT-5 and Sonnet 4.5 Retirement Dates
ElevenLabs added gpt-6.1-terra as an ElevenAgents agent and analysis LLM on the standard and fast service tiers, with reasoning effort from none to max (default medium), and new agents created without an explicit LLM now use gemini-3.5-flash instead of gemini-2.5-flash, with the default backup LLM cascade led by Gemini 3.5 Flash models; existing agents are unchanged. gpt-4.1-nano and gpt-4o-2024-05-13, which OpenAI shut down on October 23, are deprecated immediately, and agents using them now run on gpt-5.6-luna and gpt-5.6-sol respectively. gpt-5, gpt-5-mini, gpt-5-nano and their 2025-08-07 snapshots are deprecated ahead of OpenAI’s December 11 shutdown, with traffic moving to gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna in steps from 10% on November 11 to 100% on December 6, and the List LLMs endpoint now returns deprecation_info for these models. Claude Sonnet 4.5 is deprecated on November 30 with Claude Sonnet 4.6 as its replacement, and the fallback ramp for gemini-2.5-flash and gemini-2.5-flash-lite now starts on November 8 instead of November 1, keeping the November 30 cutoff. Source
3. ElevenLabs Adds an Analysis Items API, Realtime Transcript Redaction and a First Unity SDK for Voice Agents
ElevenLabs now exposes ElevenAgents evaluation criteria and data collection items as reusable analysis items under /v1/convai/analysis-items, where they can be created, listed, updated, copied and deleted, attached to or detached from many agents and branches in one request, versioned, and checked for dependent agents before a change; inline criteria in agent configs keep working and are stored as attached items. Realtime Speech to Text can now redact entities: passing entity_redaction and entity_redaction_mode as query parameters on the realtime WebSocket (in the batch transcription format) makes each committed transcript arrive with a following redacted_transcript event carrying the original and redacted text, supported from the Python and JavaScript SDKs v2.76.0 and @elevenlabs/client 1.34.0. Enterprise voice agents can enable noisy environment detection through the API and CLI with conversation_config.turn.noisy_environment_detection, so the agent speaks a configurable noisy_environment_message once when it hears other people in the caller’s background. The entry also ships Unity SDK v0.1.0 for voice agents (Unity 6.3 LTS or later, running in the Editor, desktop players and WebGL) and raises the Speech Engine conversation limit to 2 hours from 40 minutes. Source