ElevenLabs API
Programmatic access to the most realistic AI audio platform. Text to speech, speech to text, voice cloning, dubbing, sound effects, music generation, and conversational AI agents, all via REST API, WebSocket, and official SDKs.
What it is
The ElevenLabs API provides developer access to the full suite of ElevenLabs audio AI models and capabilities through REST endpoints, WebSocket streaming, and official Python and TypeScript SDKs (currently at version 39.x). It powers text to speech (TTS) with multiple models including the latest Eleven v3 (70+ languages, high emotional range), Multilingual v2 (29 languages, most stable for long form), and Flash v2.5 (ultra low latency at ~75ms). Beyond TTS, the API covers speech to text transcription (Scribe v2, 90+ languages), speech to speech voice changing, voice cloning (instant and professional), automatic dubbing, sound effects generation, music generation (with song_id for referencing and editing generated tracks), voice isolation, and conversational AI agents via the ElevenAgents platform. The WebSocket API now supports multimodal_message events, enabling agents to send and receive text, images, and other media types alongside audio in real time conversations. Authentication is via API key, and all endpoints support both synchronous HTTP and WebSocket streaming for real time applications. The Environment Variables API allows developers to manage configuration and secrets for their agents programmatically. The API uses a flat per unit pricing model with credits included in subscription tiers (Free through Enterprise) and a pay as you go option. Flash and Turbo models cost $0.05 per 1K characters while Multilingual v2/v3 cost $0.10 per 1K characters, with rates consistent across all tiers. SDKs are available for Python, TypeScript/JavaScript, React, Swift, and Unity (C#). The API is used in production by Disney, Epic Games, NVIDIA, Deutsche Telekom, and thousands of other companies. It is the backbone for building voice AI products, from audiobook platforms to real time phone agents.
Who it is for
Developers building voice enabled applications, chatbots, or AI agents that need the most natural sounding text to speech API available
Companies building real time conversational AI phone agents, customer support bots, or multimodal agents (text, images, and audio) with the ElevenAgents WebSocket API
Content platforms and SaaS products adding voiceover, narration, or read aloud features for their users
Game developers integrating dynamic character dialogue, procedural voice acting, and custom sound effects via API
Media companies and localization teams automating video dubbing across 29+ languages while preserving original speaker voices
Podcast and audiobook platforms generating AI narrated content at scale with cloned or designed voices
Accessibility teams adding high quality speech output to applications for visually impaired users
Pros and cons
Pros
- +The most natural sounding TTS API available. Eleven v3 and Multilingual v2 produce speech that is nearly indistinguishable from human voice in most cases.
- +Comprehensive API surface: TTS, STT, voice cloning, dubbing, sound effects, music, voice changer, voice isolator, and conversational AI agents all accessible from a single API key.
- +Ultra low latency WebSocket streaming (~75ms with Flash v2.5) makes it suitable for real time conversational applications and live voice agents.
- +Official SDKs for Python, TypeScript, React, Swift, and Unity (C#) with clean, well documented interfaces and rapid release cadence (versions 37 through 39 shipped in recent months). Getting started takes minutes.
- +Flexible model selection: choose between highest quality (v3/Multilingual v2), lowest latency (Flash v2.5), or balanced (Turbo v2.5) depending on your use case.
- +Voice cloning from a single 30 second audio sample gives developers programmatic access to create custom voices for their applications.
- +Flat per unit pricing across all tiers plus a pay as you go option means you can start free and scale predictably. Flash models cost $0.05 per 1K characters, half the rate of premium models at $0.10.
Cons
- −Character based pricing gets expensive at high volumes. TTS costs a flat $0.10 per 1K characters for Multilingual v2/v3 across all tiers, which is significantly more than Amazon Polly or Google Cloud TTS.
- −Eleven v3 (the best model) has a 5,000 character limit per request, requiring text splitting for longer content. Flash and Turbo models allow up to 40,000 characters.
- −Concurrency limits are relatively low on lower tiers (2 concurrent Multilingual v2 requests on Free, 3 on Starter). High traffic applications need Scale tier or above.
- −Credits do not roll over beyond two months, and unused credits expire if you downgrade or cancel your subscription.
- −Text normalization (numbers, dates, currencies) is disabled by default on Flash v2.5 to maintain low latency. Enabling it requires Enterprise plan or pre processing text before sending to the API.
- −ElevenAgents (Speech Engine) pricing at $0.08 per minute is now public, but burst pricing ($0.16/min) applies during spikes. Enterprise pricing for high concurrency voice agents still requires contacting sales.
Pricing
- 10,000 credits per month (~10 min TTS with Multilingual v2)
- ~20 min TTS with Flash models
- 4.5 hours speech to text
- 30,000 credits per month (~60 min Multilingual, ~120 min Flash)
- 27 hours speech to text
- Instant voice cloning via API
- 121,000 credits per month (~220 min Multilingual, ~440 min Flash)
- 100 hours speech to text
- Professional voice cloning
- 192kbps audio quality via API
- 600,000 credits per month (~990 min Multilingual, ~1,980 min Flash)
- 450 hours speech to text
- 44.1kHz PCM audio output via API
- 1,800,000 credits per month (~2,990 min Multilingual, ~5,980 min Flash)
- 1,359 hours speech to text
- 3 workspace seats included
- 6,000,000 credits per month (~9,900 min Multilingual, ~19,800 min Flash)
- 4,500 hours speech to text
- 10 workspace seats included
- Custom credit allocation
- Elevated concurrency limits
- Custom SSO and admin controls
Prices change; check the official pricing →