ElevenLabsBetaVerified 2026-10-30

ElevenLabs Eleven v4 Turbo (text to speech)

ElevenLabs’ real-time text-to-speech model for voice agents (model_id eleven_v4_turbo), released 2026-09-28. ElevenLabs states a median latency of about 100 ms, excluding application and network latency (its 2026-10-01 launch post on X says ~150 ms). API price $0.04 per 1,000 characters, 90+ languages.1

CompareDocs

PricingUSD, as published

ItemValue
List price$0.04 per 1K characters2
Free tierYes

API: $0.04 per 1,000 characters on every plan (ElevenLabs shows this as about $0.04 per minute of audio). Plans: Free, Starter $6/month, Creator $22/month, Pro $99/month, Scale $299/month, Business $990/month; included v4 Turbo characters per plan are 20,000, 150,000, 550,000, 2,475,000, 7,475,000 and 24,750,000.2 [Verified 2026-10-30 on https://elevenlabs.io/pricing/api]

Not in the cost calculator: Billed per character of text ($0.04 per 1,000 characters), which the calculator’s token inputs cannot express.

LimitsPer request

LimitValue
Context windowNot published
Rate limit14–105 WebSocket sessions

Text to Dialogue WebSocket sessions (each open connection holds one): Free 14, Starter 21, Creator 35, Pro 70, Scale 105, Business 105, Enterprise elevated. A new connection beyond the limit is rejected with too_many_concurrent_requests.4 [Verified 2026-10-30 on https://elevenlabs.io/docs/models]

ElevenLabs meters concurrency rather than requests per minute and publishes no fixed requests-per-minute limit. Responses carry current-concurrent-requests and maximum-concurrent-requests headers.4 [Verified 2026-10-30 on https://elevenlabs.io/docs/models]

Commercial use: paid plans get commercial rights to generated audio; the free plan is non-commercial with attribution. Zero Retention Mode is for select Enterprise customers only.56 [Verified 2026-10-30 on https://elevenlabs.io/docs/overview/administration/billing and https://elevenlabs.io/docs/eleven-api/resources/zero-retention-mode]

StatusBeta

  • Available since 2026-09-28 (changelog), with no alpha or beta label on the models page or changelog.1 [Verified 2026-10-30 on https://elevenlabs.io/docs/changelog/2026/9/28]
  • Latency: a median of about 100 ms, excluding application and network latency, per ElevenLabs. The @elevenlabsdevs launch post says ~150 ms. Both are ElevenLabs’ own figures.47 [Vendor-reported, checked 2026-10-30 on https://elevenlabs.io/docs/models and https://x.com/elevenlabsdevs/status/2105721730756931930; not reproduced by AgentAPIList]

Auth

MethodAPI key (xi-api-key)
DetailsAPI key sent in the xi-api-key header (or Authorization). The Text to Dialogue WebSocket also accepts a single_use_token query parameter for browsers; anonymous sessions are rejected.3 [Verified 2026-10-30 on https://elevenlabs.io/docs/api-reference/text-to-dialogue/ttd-websocket]

FeaturesFrom lib/apiFeatures.js

YesNoN/ANot published
CapabilityStateNote
Free tier20,000 chars/month, non-commercialYes220,000 chars/month, non-commercial
StreamingText to Dialogue WebSocketYesText to Dialogue WebSocket
Official SDKsPython, TypeScriptYesPython, TypeScript
BatchUnverified
Structured JSON outputUnverified
Tool / function callingUnverified
MCP supportUnverified
Fine-tuningUnverified
Open source / self-hostUnverified
SLA publishedUnverified
EU / data residencyUnverified

Fit and safety

Good for

  • Good for voice agents and interactive apps where you bring your own LLM and speech recognition and need expressive, low-latency speech in many languages.

Watch out

  • Per-character billing, so cost grows with what the agent says, not with call length. SSML is not supported on Eleven v4.

How it works

  1. Model ID eleven_v4_turbo, the real-time variant of Eleven v4 (eleven_v4 is the higher-quality variant at $0.08 per 1,000 characters). Same model family, 90+ languages, voice cloning (Instant and Professional Voice Clones) and audio tags such as [whispering] or [laughing].
  2. Streaming: ElevenLabs documents eleven_v4_turbo for the Text to Dialogue WebSocket (GET /v1/text-to-dialogue/stream-input). One voice ID per v4 session; text is buffered until about 40 characters and 8 words, then partial audio chunks stream back. flush, close_socket and keep_alive control messages; idle sockets close after 20 seconds.
  3. Voice settings: Stability and Similarity, plus Clone Strength for Professional Voice Clones. Style and Speed sliders are not available and SSML is not supported on Eleven v4.
  4. Official SDKs: Python and TypeScript for the REST API; the ElevenAgents SDKs also cover React, React Native, JavaScript, Kotlin, Swift and Unity.
  5. Since 2026-10-05, eleven_v4_turbo is the default TTS model for ElevenAgents conversational agents (was eleven_flash_v2).

Compare

Put ElevenLabs Eleven v4 Turbo (text to speech) side by side with other APIs on price, limits and features. Open the comparison table →

SourcesFootnotes

  1. 1ElevenLabs changelog: 2026-09-28elevenlabs.io/docs/changelog/2026/9/28Verified 2026-10-30
  2. 2ElevenLabs API pricingelevenlabs.io/pricing/apiVerified 2026-10-30
  3. 3ElevenLabs API reference: Text to Dialogue WebSocketelevenlabs.io/docs/api-reference/text-to-dialogue/ttd-websocketVerified 2026-10-30
  4. 4ElevenLabs docs: Modelselevenlabs.io/docs/modelsVerified 2026-10-30
  5. 5ElevenLabs docs: Billingelevenlabs.io/docs/overview/administration/billingVerified 2026-10-30
  6. 6elevenlabs.io/docs/eleven-api/resources/zero-retention-modeelevenlabs.io/docs/eleven-api/resources/zero-retention-modeVerified 2026-10-30
  7. 7@elevenlabsdevs launch post on Xx.com/elevenlabsdevs/status/2105721730756931930Checked 2026-10-30
  8. 8ElevenLabs docs: Eleven v4elevenlabs.io/docs/overview/capabilities/text-to-speech/eleven-v4Cited
  9. 9ElevenLabs changelog: 2026-10-05elevenlabs.io/docs/changelog/2026/10/5Cited