ElevenLabs opens Dubbing v2 to its API and adds guardrail-tracking fields to ElevenAgents conversation logs; changelog names no insurance terms

Voice-AI company ElevenLabs made its Dubbing v2 translation tool available through its API and added new fields for tracking automated guardrails inside ElevenAgents, its voice-agent product, according to an Aug. 10 changelog entry.

"Dubbing v2 is now available through the API," the changelog states, adding that "the new project-based API keeps source transcripts and translations as editable JSON." Developers can "create a project from a file or URL, add one or more target languages, edit individual transcript segments or translations, then regenerate only the regions that changed," rather than reprocessing an entire translation after a single edit.

The same update expanded what ElevenAgents records about its own guardrails — the automated checks ElevenLabs says let customers "define guardrails, run simulations, and validate agent behavior with automated tests" before an agent goes live. New optional parameters let developers filter a list of past conversations by the type of guardrail applied. On individual calls, "transcript messages add triggered_guardrails," recording which guardrail fired, and "conversation turns add optional producing_llm," identifying which model generated a given response. A separate "branch divergence status" change lets developers compare agent-configuration branches before a merge; ElevenLabs' JavaScript SDK release notes describe the update overall as having "added conversation guardrail filters, agent branch commit status, agent alerting and MCP requestMeta support."

ElevenLabs says it has "4M+ agents deployed and delivering results" across ElevenAgents, according to the company's own product page as captured in AI Dispatch's index. The Aug. 10 changelog entry does not say how long guardrail-trigger data is retained or who can access it after a call ends; it names no insurance carrier, coverage limit or indemnity term.