What is Eleven v4 Turbo?
Eleven v4 Turbo is the real-time sibling of Eleven v4. It shares v4's new architecture, audio tags and voice cloning, but is tuned for speed: ElevenLabs puts its median inference latency at about 100 ms and its time to first speech at about 150 ms. You stream text in as your language model writes it and get audio back before the sentence ends, through the Text to Dialogue WebSocket or inside ElevenAgents.
It succeeds Eleven v3 Conversational as ElevenLabs' expressive option for live agents, at the same $0.04 per 1,000 characters. For recorded audiobooks, games and multi-voice scenes, Eleven v4 is the better fit.
What changed from Eleven v3 Conversational?
Much lower latency and more languages on the new v4 architecture, at the same price per character.
- Latency
- About 100 ms median inference, against about 280 ms for v3 Conversational.
- Languages
- 90+ languages, against 70+.
- Voices
- Professional Voice Clones work the same as on Eleven v4, and older clones need retraining.
- Controls
- Only Stability and Similarity settings; no Style or Speed sliders and no SSML.
- Price
- Both cost $0.04 per 1,000 characters on the API.
Who is Eleven v4 Turbo best for?
Eleven v4 Turbo is best for voice agents, AI assistants and interactive characters that need expressive speech with a short wait, in 90+ languages.
Customer support voice agents
It answers in about 150 ms to first speech, and ElevenLabs built it together with its ElevenAgents platform.
AI assistants that should sound human
It keeps Eleven v4's emotional range, so confirmations, apologies and holds do not all come out in the same tone.
Interactive game characters
Audio tags such as [laughs] or [whispers] work in real time, so characters can react as a player speaks.
One brand voice in many languages
Professional Voice Clones work the same as on Eleven v4, and a cloned voice speaks 90+ languages with a native accent.
What's included with Eleven v4 Turbo?
Bidirectional streaming, real-time audio tags, instant and professional voice cloning, native accents across languages, ElevenAgents integration and session-based streaming limits.
- Bidirectional streaming
- Send text as your LLM generates it and receive audio chunks back before the sentence is finished.
- Audio tags
- Square-bracket tags set emotion, reactions and delivery in real time, the same as on Eleven v4.
- Voice cloning
- Instant Voice Clones from as little as 10 seconds of audio, and Professional Voice Clones shared with Eleven v4.
- Native accents across languages
- A voice recorded in one language speaks the others fluently with a local accent instead of its original one.
- ElevenAgents
- Runs inside ElevenLabs' agent platform, which bills calls by the minute and adds text messages, telephony and a knowledge base.
- Session-based streaming limits
- Each open WebSocket counts as one dialogue session from its own pool, from 14 at a time on Free to 105 on Scale and Business.
How much does Eleven v4 Turbo cost?
Eleven v4 Turbo costs $0.04 per 1,000 characters on the ElevenLabs API; the free plan includes about 20 minutes of it a month, and paid plans start at $6 a month.
- 20,000 characters of v4 Turbo a month on the API, about 20 minutes
- 15 minutes of ElevenAgents calls
- Personal use only
- Half the price of Eleven v4
- Streams over the Text to Dialogue WebSocket
- One voice per streaming connection
- 75 minutes of ElevenAgents calls, 6 at once
- Commercial license
- Instant voice cloning
- 275 minutes of ElevenAgents calls, 10 at once
- Professional voice cloning
- Extra call minutes at $0.08
In ElevenAgents, calls beyond a plan's included minutes cost $0.08 a minute, or $0.16 when burst pricing lets you exceed your concurrency limit, and the language model is billed on top; ElevenLabs adds no telephony fee. Pro is $99 a month (1,238 call minutes), Scale $299 (3,738) and Business $990 (12,375); Enterprise is priced on request. On the API, each plan's included characters follow the per-character rate. Yearly billing gives two months free. Prices are in US dollars and exclude tax. Full pricing on elevenlabs.io. We checked these prices on .
How do readers rate Eleven v4 Turbo?
No one has rated Eleven v4 Turbo yet. Be the first.
Sign in to rate. One rating per account.
Eleven v4 Turbo pros and cons
It brings Eleven v4's expressive voices to live calls at about 150 ms to first speech and half v4's price, but it streams one voice per connection, keeps v4's control limits and costs more per minute than Google's option.
Pros
- Fast enough for live calls: about 100 ms median inference and about 150 ms to first speech, which ElevenLabs measured as quicker than Cartesia Sonic 3.6 and OpenAI's GPT-4o mini TTS.
- Keeps Eleven v4's expressive range and audio tags, where ElevenLabs' older fast Flash models trade expression for speed.
- Half the price of Eleven v4 at $0.04 per 1,000 characters, the same as the Eleven v3 Conversational model it succeeds, with 90+ languages against 70+.
- Professional Voice Clones work the same as on Eleven v4, so one brand voice can cover recorded content and live calls.
Cons
- Streams one voice per connection: multi-speaker scenes over the WebSocket need Eleven v4, which takes up to 10.
- Not the fastest ElevenLabs model: Flash v2.5 answers in about 75 ms at the same $0.04 per 1,000 characters, with less expression and 32 languages.
- Same control limits as Eleven v4: only Stability and Similarity settings, no Style or Speed sliders, no SSML, and clones made before v4 must be retrained.
- Costs about twice as much per minute as Gemini 3.8 Flash-Lite TTS at Google's paid rate, which also has a free API tier.
What are the best Eleven v4 Turbo alternatives?
Eleven v4 for recorded, multi-voice audio, Gemini 3.8 Flash-Lite TTS for the lowest cost per minute, and Voxtral TTS for a lower price per character.
Eleven v4FreeChoose it when the audio is recorded rather than live: it is ElevenLabs' highest-quality version, voices up to 10 speakers per dialogue connection and takes 10,000 characters per request, at $0.08 per 1,000 characters.
Gemini 3.8 Flash-Lite TTSFreeChoose it when cost per minute matters most for a voice agent: Google's paid rate works out to about $0.018 a minute of audio, with a free API tier and 101 languages.
Voxtral TTSFreeChoose it when price per character matters and your languages are covered: Mistral charges $0.016 per 1,000 characters, 40% of v4 Turbo's rate, and reports 70 ms model latency in 9 languages.
Pocket TTSFreeChoose it when you want a 100M text-to-speech model with voice cloning that runs in real time on a CPU.
Questions about Eleven v4 Turbo
What is the difference between Eleven v4 Turbo and Eleven v4?
Eleven v4 is tuned for recorded content where quality matters most and costs $0.08 per 1,000 characters. Eleven v4 Turbo is the low-latency version for agents at $0.04, with about 100 ms median inference latency and one voice per streaming connection.
Who is Eleven v4 Turbo for?
Eleven v4 Turbo is best for voice agents, AI assistants and interactive characters that need expressive speech with a short wait, in 90+ languages.
Is Eleven v4 Turbo free?
Yes. Eleven v4 Turbo has a free plan: 20,000 characters of v4 Turbo a month on the API, about 20 minutes.
How much does Eleven v4 Turbo cost?
Eleven v4 Turbo's paid plans are API, pay as you go at $0.04 per 1,000 characters, Starter at $6 per month and Creator at $22 per month.
What are the best Eleven v4 Turbo alternatives?
The closest alternatives on Recaply are Eleven v4, Gemini 3.8 Flash-Lite TTS, Voxtral TTS and Pocket TTS.
Where can I try Eleven v4 Turbo?
On its official website, elevenlabs.io.
Eleven v4 Turbo or Flash v2.5?
Both cost $0.04 per 1,000 characters on the API. Flash v2.5 is quicker at about 75 ms and covers 32 languages; Eleven v4 Turbo takes about 100 ms but is far more expressive, follows audio tags and covers 90+ languages.
Is Eleven v4 Turbo faster than Cartesia Sonic 3.6?
By ElevenLabs' own test, yes: it measured median time to first speech at about 150 ms for Eleven v4 Turbo against 262 ms for Sonic 3.6, with network latency removed. Cartesia quotes sub-90 ms latency for Sonic, so test both in your own setup.
How do I stream Eleven v4 Turbo?
Open the Text to Dialogue WebSocket with model_id eleven_v4_turbo and register exactly one voice. The server starts sending audio once it has about 40 characters and 8 words, and you can send flush to force it sooner.
Can I use Eleven v4 Turbo commercially?
Only on a paid plan. ElevenLabs' Free plan is for personal use; the Starter plan at $6 a month adds a commercial license, along with 75 minutes of ElevenAgents calls, and every higher plan keeps it.
Our verdict
Eleven v4 Turbo is the ElevenLabs voice to pick for live agents that should sound human rather than read aloud.
It keeps Eleven v4's expressive range and voice clones at about 150 ms to first speech, for half v4's price per character.
- Skip it for
- Recorded audiobooks and multi-voice scenes, where Eleven v4 fits better, and very high-volume agents, where Google's Flash-Lite TTS costs less per minute.
- Check first
- Measure latency end to end with your own language model and network, since ElevenLabs' figures leave both out.
ElevenLabs Voice Design V3
Scribe v2 Realtime
Chatterbox Turbo
Eleven Realtime
Eleven v3
Scribe v2