ElevenLabs AI Updates: April 21, 2026

1. ElevenAgents Adds Workflow Node Overrides, Speaker Role Detection, and New LLM Options

ElevenLabs. ElevenLabs shipped a broad ElevenAgents update letting conversations start at a specific workflow node via a new starting_workflow_node_id parameter, added a configurable analysis_llm field for post-call analysis (default gemini-2.5-flash), and exposed new LLM options including a gpt-5.4 provider and an xhigh reasoning effort tier. Speech-to-Text gained a detect_speaker_roles parameter that returns “agent” and “customer” labels (with a 10% cost surcharge) instead of generic speaker numbers. Workspaces also picked up a credit-usage analytics endpoint broken down by product type, and the JS, React, React Native, and widget SDKs were updated to fix conversation startup readiness and expose a new onConversationCreated callback. Source

2. ElevenLabs Launches YouTube Channel for Developer Education

ElevenLabs. Alongside the April 20 product update, ElevenLabs announced a new YouTube channel featuring tutorials, API walkthroughs, and technical deep dives aimed at helping developers ramp on the platform’s voice, agent, and speech APIs. The launch coincides with API documentation expansions including newly published 2,000-character guidance for Text to Dialogue requests. Source