Eleven Labs Release Notes

ElevenAgents

• Structured procedures compile on save and publish: Update agent and Create agent draft now compile and validate structured procedures automatically. Integrations no longer need to call the compile endpoint or manually update the generated workflow. Invalid procedures return per-procedure errors as 400 procedure_validation_failed. Agents without structured procedures are unaffected.
• Compile procedures endpoint is legacy: Compile procedures is marked legacy and should not be used by new integrations. It remains available for existing callers.
• Merge proposals: Merge proposals are available in alpha. Teams can open a proposal for an agent branch, review configuration changes and test results, leave comments or approval decisions, and merge approved changes into a target branch. See Merge proposals.

SDK releases

JavaScript SDK …

Eleven v4 and Eleven v4 Turbo

Eleven v4 and Eleven v4 Turbo are now available. Eleven v4 provides higher-quality, more expressive speech and improved voice cloning across more than 90 languages. Eleven v4 Turbo provides the same model family for real-time applications, with median inference latency of approximately 100 ms.

Use eleven_v4 through the Text to Dialogue API for content creation and long-form audio. Use eleven_v4_turbo through the Text to Dialogue WebSocket for agents and interactive applications.

Speech to Text

• Transcript editing: Batch and realtime transcription now accept a natural-language edit instruction of up to 2,000 characters. Batch responses return the result in edited_transcript, while realtime sessions emit an edited_transcript event. See the batch and realtime guides for incompatibilities, response handling and pricing.

ElevenAgents

• Agent models: Added gpt-6-sol, gpt-6-luna, glm-52, deepseek-v41-flash, claude-opus-5 and claude-opus-5-5 to the supported agent LLM options.
• Agent test usage: Test invocation summaries can include total credits and USD price. Individual test runs can include credits and a charging breakdown.

Flows …

ElevenAgents

• Parallel tool calls: Agent prompt configuration now includes enable_parallel_tool_calls (boolean, default true). When enabled, supported models can execute multiple tools in one turn.
• Agent models: Added gpt-6-astra to the agent LLM options.
• Alerting: Agent alerting now supports Slack channels alongside PagerDuty and webhook notifiers.
• Phone number search: Added a cursor-paginated phone number endpoint with filters for provider, outbound support, assigned agent and branch, label and phone number.

Image and video

• GPT Image 2.5 models: Image generation now supports gpt-image-2.5-flare and gpt-image-2.5-sunburst. Both models accept up to 10 reference images, quality levels through max, 14 fixed aspect ratios plus auto, and 1K, 2K or 4K output.

Speech Engine

• Cascade timeout: Speech Engine create and update requests now accept cascade_timeout_seconds (number, 2–15, default 4) to control how long ElevenLabs waits for an upstream speech engine before retrying.

Text to Speech

• Multi-context turn boundaries: The is_final_audio_for_turn marker is now emitted after every buffered byte for that turn, including MP3 and Opus output. Clients can use the marker as an exact turn boundary without switching to PCM.

SDK releases

JavaScript SDK …

ElevenAgents

• Call queueing and hold audio: Added call queueing for agents at their concurrency limit. Queued callers hear hold audio and receive queue_status events until they are admitted or time out. New API methods upload or delete custom hold audio.
• Test version metadata: Agent test results now identify the agent version they ran against and whether they included draft changes. Invocation summaries also indicate when partial resubmissions ran against different versions.
• Model and account metadata: Agent LLM configuration now accepts gemini-3.8-flash. WhatsApp account responses identify whether the account uses the Cloud API or coexistence signup flow.

Music

• Music v2.5 API support: Music endpoints and SDKs now accept music_v2_5. Composition plan generation chunks support up to 6,132 characters, with up to 30 lines of 200 characters each.

SDK releases

JavaScript SDK

• v2.68.0 - Added methods to upload and delete agent hold audio. Regenerated types for call queueing, procedure references, draft branch creation, RAG query limits, test version metadata, gemini-3.8-flash and music_v2_5.

Python SDK …

ElevenAgents

• Workspace-wide conversation tickets: Added List workspace tickets to retrieve conversation triage tickets across accessible agents. Filter by status or assignee and use cursor pagination to retrieve up to 100 tickets per page.
• Dynamic variable conversation filters: List conversations and Text search conversation messages now accept repeatable dynamic_variable_params filters. Use name:op:value with eq, gt, gte, lt or lte; comparison operators require numeric values.
• Twilio answering machine detection: Outbound telephony and batch call configurations now support optional twilio_machine_detection. Choose enable for an early human-or-machine verdict or detect_message_end to wait for a voicemail greeting to finish. Results arrive through the new answering_machine_detection webhook event.
• Data collection value constraints: Data collection properties can now use allowed_values to name a dynamic variable containing the permitted values. The previous allowed_values_dynamic_variable field is deprecated.
• Agent metadata: List agent branches responses now indicate when a draft was created and whether it predates the branch tip. Agent and Speech Engine summaries now require voice_id. …

ElevenAgents

• Phone keypad input: Agent conversation configuration now supports optional dtmf_input_settings for DTMF input. Configure the digit timeout, use # as a terminator and redact keypad entries from transcripts, logs and analysis.
• Conversation attachment history: Get conversation transcript entries now include file_inputs, an array containing every attached file's ID, name, MIME type and URL.
• Agent tag filtering: List agents now accepts an optional tags query parameter. Repeat the parameter to match agents with any of the supplied tags.
• Batch campaign context: Get conversation now identifies batch calls with nullable campaign metadata containing the required campaign_id and campaign_lead_id.

Music

• Waveform visual data: Generate music detailed, Stream music with details and Upload music now accept optional with_waveform_visual (boolean, default false). Opted-in responses return a low-resolution waveform_visual integer array with four samples per second.

Dubbing …

ElevenLabs CLI

The ElevenLabs CLI v1.0.0 is now available. Every ElevenLabs API operation is available as a subcommand, with JSON, table, YAML and CSV output, automatic pagination and shell completion.

The CLI also provides local configuration workflows for ElevenAgents. Store agent, tool and test configurations as files, synchronize them with push and pull commands, manage branches, run tests, install ElevenLabs UI components and select a data residency region.

Procedures

Procedures are now generally available in ElevenAgents. A procedure contains task-specific instructions and a trigger that determines when they apply. During a conversation, the agent loads the relevant procedure, allowing one agent to handle distinct tasks without placing every instruction in its system prompt.

Use free-form procedures when the agent can adapt the wording or order of instructions. Use structured procedures when steps must run in a defined order. Both types can be used alongside workflows on the same agent.

ElevenAgents

• Conversation triage tickets: Added APIs for creating tickets about agent performance, creating manual follow-up tickets, listing tickets, assigning workspace members, updating status and assignee, and adding ticket-level or turn-level comments. …

MCP server

• Local MCP server deprecated: The local ElevenLabs MCP server and MCP player are deprecated in favor of the hosted MCP server. Both repositories are archived and will no longer receive updates. The hosted server is available at https://api.elevenlabs.io/v1/mcp, signs in with your ElevenLabs account through OAuth and requires no local installation or API key. See the hosted MCP server documentation for setup instructions for Claude, Cursor and other MCP clients.

ElevenCreative

• Image, video and speech generation APIs: New asynchronous Flows APIs create, list and retrieve video, image and speech generations. Generation responses expose pending, generating, completed and failed states, and requests can include a webhook target for completion events.
• Reusable media assets: New asset APIs upload, list, retrieve and delete workspace media. Image, video and audio references can use an uploaded asset, a previous generation or inline base64 data, allowing generation jobs to be chained. See References and assets.

ElevenAgents

• Lightweight conversation summaries: Get conversation summary returns the generated title, transcript summary, call outcome and plain user or agent messages without loading the full transcript. The optional max_messages query parameter is an integer from 1 to 200 (default 40); longer conversations set messages_omitted to true. Useful for agents that need to preserve context.
• Conversation topic pagination and sorting: Get agent conversation topics adds optional cursor pagination and sorting. Use page_size (integer, 1–100), cursor (string), sort_by (conversations, sentiment or success_rate) and sort_direction (asc or desc). Responses add has_more and nullable next_cursor. …

Dubbing v2 API

Dubbing v2 is now available through the API. It translates audio and video into more than 90 languages while preserving each speaker's voice, tone and pacing.

The new project-based API keeps source transcripts and translations as editable JSON. 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. Follow the Dubbing quickstart to create your first dub.

ElevenAgents

• Conversation guardrail filters: List conversations adds optional guardrail_types (GuardrailType[]) and custom_guardrail_names (string array) query parameters. The existing conversation hierarchy can now be filtered with optional parent_conversation_id (string).
• Branch divergence status: List agent branches adds optional include_commit_status (boolean, default false). When enabled, branch summaries include nullable commits_ahead, commits_behind and merged_into_branch_id fields.
• Conversation and tool controls: Client overrides now support max_duration_seconds and pronunciation_dictionary_locators. MCP server configuration adds request_meta for passing MCP _meta values to tool calls, and webhook tool timeouts now support up to 300 seconds. …

ElevenAgents

• Test-specific tool mocks: Agent simulations and unit tests can now define tool_mock_overrides, keyed by tool ID, to replace shared mocks for one test. ToolResponseMockConfig also adds optional is_error (boolean, default false) so a mock can exercise tool-failure paths.
• Translated workflow messages: Literal Say node messages add text_translations, a map of language keys to TranslatedString objects. The field is optional on input and required on output, so strict response validators must accept it.
• Conversation version filters: List conversations and Text search conversation messages add the optional version_id query parameter (string) to filter results by agent version.
• Knowledge base system tool: Agent configuration adds a knowledge_base system tool with configurable enabled_strategies (SearchStrategy[]). Agent knowledge base RAG queries add optional use_agent_defaults (boolean, default true); set it to false to use neutral retrieval settings while retaining the agent's embedding model.
• Agent voice activity detection: Agent create, update, response and workflow-override schemas add vad configuration. VADConfig currently supports background_voice_detection (boolean, default false).
• WebSocket custom LLMs: CustomLLMAPIType adds websocket. …

ElevenAgents

• WhatsApp branch and environment overrides: Outbound WhatsApp messages and outbound calls now honor branch_id and environment passed in conversation_initiation_client_data. Overrides are validated before anything is sent — an unknown branch or environment is rejected with an error — and persist for the conversation, so a customer's reply resumes on the requested branch and environment. See the WhatsApp guide for examples.
• Procedure APIs: Added eight public, branch-scoped operations for managing procedures. You can create, list, retrieve and remove procedures; create, retrieve or discard draft changes; and compile procedure drafts into an agent workflow. Procedure creation accepts name, content, type and trigger. An empty trigger creates a sub-procedure, while an omitted or null trigger is derived from the procedure content.
• Knowledge base crawl jobs: Added APIs to create, list, inspect and cancel crawl jobs. Create requests require url (string) and support crawl depth, page limits, URL patterns, sitemap URLs, folder placement, automatic synchronization and removal of unavailable documents. …

Music

• Music Finetunes API: Added API support for managing Music Finetunes. Five endpoints let you create a Finetune from uploaded audio, list accessible Finetunes, retrieve training status, update metadata and visibility, or delete a Finetune. All music generation SDK methods now expose finetune_id.

ElevenAgents

• Resolve external conversation references: Added GET /v1/convai/conversations/resolve to resolve a Slack message URL or Zendesk ticket URL to a conversation. Requests require agent_id and reference query parameters (strings).
• Query agent knowledge bases: Added POST /v1/convai/agents/{agent_id}/knowledge-base/rag-query for read-only RAG queries against an agent's knowledge base. The request requires query (string), accepts optional branch_id (string), and returns ranked chunks with document, chunk, text and vector-distance metadata. This endpoint is excluded from generated SDKs.
• Knowledge base content formats: Knowledge base file, text and URL responses add optional content_format (html | markdown, default html).
• Agent voice configuration: TTS override schemas add optional client-overridable model_id (TTSConversationalModel). The default for enable_phoneme_tags changes from false to true.
• Faster backup LLM cascade: The default cascade_timeout_seconds changes from 8 to 4 seconds. The allowed range remains 2–15 seconds. …

ElevenAgents

• Per-agent sentiment analysis: Platform settings add optional sentiment_analysis (SentimentAnalysisSettings) for post-call sentiment scoring configuration per agent.
• Auto-translate transcripts: Platform settings add optional auto_translate_transcript_to_app_language (boolean) to translate conversation transcripts to the viewer's app language in history.
• Nested agent transfers: Workflows add push, pop and replace operations for nested agent transfers, with enable_nesting and return_when_nested controls for returning to a parent workflow.
• Backchannel detection: Transcript user-turn models add optional ignored_as_backchannel (boolean) to mark utterances filtered as backchannels.
• Auxiliary conversation audio: Get conversation responses add required has_auxiliary_audio (boolean) alongside existing audio flags.
• MCP environment scoping: MCP server and tool routes add optional environment query parameter (defaults to production) on list tools, get tool, create tool config and update tool config endpoints.
• Tool call sound override: Tool and MCP override schemas allow tool_call_sound: "off" to silence per-tool call sounds while inheriting server defaults when unset.
• Twilio SMS routing: Twilio phone-number import requests add optional enable_sms (boolean, default on) to control inbound SMS routing during number setup. …

Music

• Stream music with details: Added Stream music with details (POST /v1/music/detailed/stream) for SSE streaming of composed audio plus detailed metadata, including composition plan, song metadata, base64 audio chunks and optional word timestamps via with_timestamps.

ElevenAgents

• Simulate conversation deprecation: Simulate conversation and Stream simulate conversation are deprecated in favor of Create test and Run tests on the agent.
• Tool interruption mode: disable_interruptions is deprecated across agent tool and MCP server schemas. Use interruption_mode (ToolInterruptionMode: allow, disable_during_tool, disable_during_tool_and_turn) instead, with per-tool overrides via interruption_mode on tool override models.
• Numeric evaluation scoring: Evaluation criteria schemas add optional scoring_mode (binary | numeric_uniform), optional max_score (integer) and optional score_instructions (string). Evaluation results add optional numeric score. Conversation analysis payloads add optional call_success_score and success_score. …

ElevenAgents

• Branch merge and rebase previews: Added Preview merged configuration (GET /v1/convai/agents/{agent_id}/branches/{source_branch_id}/merge-preview) and preview rebased configuration (GET /v1/convai/agents/{agent_id}/branches/{branch_id}/rebase-preview) to inspect merged or rebased agent configuration before committing. Both return MergePreviewResponseModel with the resulting config, overridden_fields, and source_identical_to_target.
• Branch usage metrics: List agent branches branch summary responses add optional calls_7d (integer) with seven-day call volume per branch.
• Auth connection status: Workspace auth connection response models replace OAuth-only OAuthConnectionStatus with shared AuthConnectionStatus (active, refresh_failed, revoked, credential_invalid). Bearer, Basic, mTLS, custom header, OAuth, OAuth2 JWT, Slack, URL secret and WhatsApp auth responses add optional status, status_detail and status_updated_at fields.
• Tool response filtering: HTTP and agent tool configuration schemas add optional response_filter to trim tool responses before they reach the LLM, with response_filter_mode and response_filters on related override models. …

ElevenAgents

• Branch rebase: Added POST /v1/convai/agents/{agent_id}/branches/{branch_id}/rebase to rebase a branch onto the latest main while preserving branch-specific changes. Agent version parent models add optional rebased_from_version_id (string) to track the source version after a rebase.
• Conversation product filtering: List conversations and Text search conversation messages add optional conversation_product_type query parameter (agents | speech_engine) to filter results by product. Conversation models introduce ConversationProduct for this distinction.
• Conversation termination filtering: List conversations adds optional termination_reasons query parameter (array of strings) to filter conversations by how they ended.
• Per-agent topic discovery: Create agent and Update agent platform settings add optional topic_discovery (TopicDiscoverySettings) for per-agent topic discovery configuration.
• Background sound configuration: Agent platform and conversation configuration models add optional background_sound (BackgroundSoundConfig) for ambient audio during conversations. …

Introducing Music v2

Music v2 is now available via the ElevenLabs API. Set model_id to music_v2 on Generate music, Stream music, Generate music detailed and Upload music to use the new model. Music v2 introduces chunk-based composition plans built from GenerationChunk and AudioRefChunk segments, giving you finer control over structure, pacing and arrangement than the prompt-based music_v1 flow.

Use music_v2 composition plans on generation and upload requests, or generate a plan first via the composition-plan endpoint. Request and response bodies accept either v1 or v2 plan shapes, so you can migrate endpoint by endpoint. MusicPrompt remains available for music_v1 only.

• Upload timestamps: Upload music accepts optional with_timestamps; responses may include words_timestamps.

ElevenAgents

• Run conversation evaluation: Added Run conversation evaluation endpoint (POST /v1/convai/conversations/{conversation_id}/analysis/evaluations/run) to rerun a single evaluation criterion against a completed conversation using RunConversationEvaluationsRequest. …