What it is
ElevenLabs launched Eleven v4 and Eleven v4 Turbo on September 28, 2026 (API IDs eleven_v4 and eleven_v4_turbo). Both use a new architecture and support more than 90 languages. Eleven v4 targets audiobooks, voiceovers and dubbing; v4 Turbo targets real-time voice agents, with a median inference latency of about 100 ms and about 150 ms to first speech, according to ElevenLabs.
Delivery can be directed with natural-language instructions and inline tags such as [laughs] or sound effects, and pronunciation with IPA. Instant voice clones need only 10 seconds of audio, and Professional Voice Clones are supported.
Our review
ElevenLabs has retaken the top of the voice rankings. On Artificial Analysis's speech arena, built from blind listener votes, Eleven v4 Turbo scores 1335 and Eleven v4 1329, a statistical tie for first, ahead of Alibaba's Qwen-Audio-3.1-TTS-Plus (1301), Cartesia Sonic 3.6 (1289) and Google's Gemini 3.8 Flash TTS (1278). ElevenLabs' own blind tests against Cartesia, Inworld and two Gemini models report win rates of 65% to 81%.
The odd result is that the cheaper model ranks fractionally higher. Turbo costs $40 per million characters against $80 for v4, and it is tuned for ElevenAgents with a quoted median of about 150 ms to first audible speech. The full v4 earns its price on long-form work: multi-speaker dialogue that takes the whole scene into account, inline direction tags such as [laughs] or [light rain], better IPA handling and Professional Voice Clone support.
Price is where rivals close the gap. Artificial Analysis lists Gemini 3.8 Flash TTS at $16.5 per million characters and Flash-Lite TTS at $11, so v4 costs roughly five times Flash for a 51-point Elo lead. Cartesia Sonic 3.6 at $49 is dearer than Turbo and ranks lower, which makes Turbo an easier call for voice agents than v4 is for bulk narration.
Buyers with sensitive audio should read the data terms. By default ElevenLabs preserves generation history and retains data "to enhance services"; Zero Retention Mode is limited to select Enterprise customers and does not cover voice cloning, dubbing or Studio. For a dubbing house cloning actors' voices, the most sensitive step is the one that cannot be made zero-retention.
Where it earns its place
An independent audiobook publisher releasing 20 titles a year
Narrate backlist novels without booking studio time
A 90,000-word novel is roughly 500,000 characters, about $40 per full pass on v4 at API rates. The $990 Business plan includes 12.4M v4 characters, enough for about two dozen books a month. Requests cap at 10,000 characters, so chapters are generated in chunks and stitched.
A contact-centre platform team building a voice agent on ElevenAgents
Answer inbound calls with a voice callers do not hang up on
Stream v4 Turbo over WebSocket for about 150 ms to first speech. At 3,000 characters of agent speech per call, 100,000 calls a month is 300M characters: about $12,000 on Turbo against about $4,950 on Gemini 3.8 Flash TTS. Scale and Business plans cap WebSocket sessions at 105.
A games studio localising dialogue into a dozen languages
Keep each character's voice consistent across languages
Clone the original actor and generate each language; ElevenLabs says clones take on a native speaker's accent while keeping the voice's identity. Get written consent and check the contract first, since voice cloning sits outside Zero Retention Mode.
Before you commit
- Setup
- API IDs eleven_v4 and eleven_v4_turbo; v4 runs on the HTTP Text to Dialogue API and Turbo on the dialogue WebSocket. A free account includes 10,000 v4 characters a month for testing.
- Integrations
- ElevenLabs API and SDKs, ElevenAgents for voice agents and ElevenCreative for production work.
- Data and privacy
- History is preserved by default and data is retained "to enhance services"; deleted items stay in backups for up to 30 days. Zero Retention Mode is for select Enterprise customers and excludes voice cloning, dubbing and Studio. Enterprise adds SSO, SOC 2 compliance, custom DPAs and HIPAA BAAs.
- Cost at scale
- API rates are $0.08 per 1,000 characters for v4 and $0.04 for Turbo on every plan, so 1M characters is $80 or $40. Plans from Starter ($6, 75,000 v4 characters) to Business ($990, 12.4M) bundle allowances at the same rates.
- Limits
- 10,000 characters per v4 request (about 10 minutes of audio). Turbo WebSocket sessions are capped from 14 on Free to 105 on Scale and Business, and close after 20 seconds idle without keep_alive. No per-request limit is published for Turbo.
What's good
- Tied first and second on Artificial Analysis's speech arena (1335 and 1329)
- Turbo at $40 per million characters outranks the pricier Cartesia Sonic 3.6
- Inline direction tags, IPA and scene-aware multi-speaker dialogue
- Instant clones from 10 seconds of audio; Professional Voice Clones supported
What's not
- v4 costs about five times Gemini 3.8 Flash TTS per character
- Zero Retention Mode is Enterprise-only and excludes cloning and dubbing
- 10,000-character cap per v4 request, so long scripts need stitching
- Closed model, no self-hosting
Best for
Audiobook, game and dubbing producers who need the most expressive voices, and voice-agent teams who want top-ranked quality at Turbo's lower rate.
Alternatives
Eleven v4 and v4 Turbo: common questions
- Is Eleven v4 and v4 Turbo worth it?
- Nothing on the public leaderboards sounds better, but Google's Gemini 3.8 Flash TTS gets within 60 Elo points for about a fifth of v4's price. We rate it 4.7 out of 5.
- How much does Eleven v4 and v4 Turbo cost?
- API: v4 $0.08 and v4 Turbo $0.04 per 1,000 characters. Subscriptions from Starter $6/month to Business $990/month include character allowances.
- What are the downsides of Eleven v4 and v4 Turbo?
- v4 costs about five times Gemini 3.8 Flash TTS per character; zero Retention Mode is Enterprise-only and excludes cloning and dubbing; 10,000-character cap per v4 request, so long scripts need stitching; closed model, no self-hosting.
- Who is Eleven v4 and v4 Turbo best for?
- Audiobook, game and dubbing producers who need the most expressive voices, and voice-agent teams who want top-ranked quality at Turbo's lower rate.
- How much does Eleven v4 and v4 Turbo cost for a team or in production?
- API rates are $0.08 per 1,000 characters for v4 and $0.04 for Turbo on every plan, so 1M characters is $80 or $40. Plans from Starter ($6, 75,000 v4 characters) to Business ($990, 12.4M) bundle allowances at the same rates.
- Is Eleven v4 and v4 Turbo safe for company data?
- History is preserved by default and data is retained "to enhance services"; deleted items stay in backups for up to 30 days. Zero Retention Mode is for select Enterprise customers and excludes voice cloning, dubbing and Studio. Enterprise adds SSO, SOC 2 compliance, custom DPAs and HIPAA BAAs.
- What are the best Eleven v4 and v4 Turbo alternatives?
- The closest alternatives we have reviewed are Cartesia Sonic 3.6 and Gemini 3.8 Flash TTS.
Answers checked Oct 22, 2026.
Make Eleven v4 and v4 Turbo? Show this rating on your site
