Genly guide

ElevenLabs v4: What's New, Pricing and v4 vs v3 for Voice-Overs

Does an AI voice finally sound natural? We put ElevenLabs v4 against v3 with our own voice-overs, and explain what's new and what it costs.

You have a solid script for your ad, but the AI voice stumbles over one word or puts the stress in the wrong place, and viewers notice. At the end of September 2026, ElevenLabs released a new generation of voices: Eleven v4. We tested it the day after launch with Dutch scripts, and here is what you need to know.

Short answer: ElevenLabs v4 is the newest text-to-speech model from ElevenLabs, with a faster variant, Eleven v4 Turbo, for live conversations. It supports more than 90 languages, and you control emotion and pace with audio tags such as [whispers] and [slowly]. Through the API, v4 normally costs $0.08 per 1,000 characters, the same as v3. In our own test on September 29, 2026, 8 out of 8 Dutch takes on v4 were correct word for word.

What is ElevenLabs v4 and what's new?

ElevenLabs published its Eleven v4 announcement on September 28, 2026. There are two models:

  • Eleven v4 is built for quality: voice-overs, video and audiobooks. A single generation can be up to 10,000 characters, twice the v3 limit.
  • Eleven v4 Turbo is the low-latency variant for voice agents and phone calls. ElevenLabs reports a median time to first speech of about 150 milliseconds.

What changed compared to v3? According to the ElevenLabs v4 page:

  • The same voice on every retake. Regenerate a line once or many times and it stays the same speaker.
  • Steady delivery on long scripts. Pacing and tone hold across a script of any length.
  • Professional Voice Clones are back. They were not supported on v3. High-fidelity cloning of your own voice works again.
  • More languages: over 90, up from 70+ on v3. See the full list in the ElevenLabs model docs.
  • Audio tags follow direction more closely. You write how a line should sound in brackets, such as [excited], [whispers], [sighs] or [pause].
Eleven v4 Eleven v4 Turbo Eleven v3
Built for Voice-overs and video, quality first Voice agents and live calls, speed first Previous generation
Languages 90+ 90+ 70+
Stands out 10,000 characters per generation, Professional Voice Clones About 150 ms to first speech 5,000 characters per generation, no Professional Voice Clones

ElevenLabs v4 vs v3: our own test

At Genly we produce voice-overs every day, mostly in Dutch, so we wanted to hear the difference ourselves. In our own test on September 29, 2026, we ran Dutch scripts through v4 and v3 with the voices Charlotte, Naomi, Daan and Ruben, then checked every take word for word.

What we tested Eleven v4 Eleven v3
Takes correct word for word 8 out of 8 3 out of 4
Notable error None "kwijt" came out as "kleid"
Length of the same sentence 10.4 seconds 9.3 seconds

What stood out:

  • Pronunciation. v4 made no mistakes in our takes. v3 turned the Dutch word "kwijt" (lost) into "kleid". In a 20-second ad, everyone hears that.
  • Pace. v4 speaks a little slower. The same sentence took just over a second longer, roughly 10 percent. If you move a script from v3 to v4, plan for that.
  • Connected scenes. When a video is built from separate scenes, you want the voice to flow as if it were one recording. In the API you pass the text before and after each scene (previous_text and next_text). That works on v4.

To be fair, this was a small test with four voices, not a guarantee. Always listen back before an ad goes live.

Based on this test, Genly has used v4 for all voices since September 29, 2026: in UGC videos, voice-overs and persona voices.

How much does ElevenLabs v4 cost?

On the API you pay per character. These are the rates on the ElevenLabs API pricing page:

Model Standard price per 1,000 characters Launch price until October 12, 2026
Eleven v4 $0.08 $0.022 (72 percent off)
Eleven v4 Turbo $0.04 $0.011 (72 percent off)
Eleven v3 $0.08 No discount
Multilingual v2 $0.08 No discount

The launch discount ran until October 12, 2026. After that, the standard rate applies. The good news: v4 is not more expensive than v3 or Multilingual v2, so switching costs you nothing extra.

A quick example. A 30-second voice-over is roughly 75 words, around 450 characters. One take on v4 then costs about 4 cents through the API. Generate ten versions to pick the best one and you still stay under 40 cents.

In the ElevenLabs app you pay with credits from your subscription. According to the ElevenLabs pricing page, there is a free plan with 10,000 credits a month, and paid plans start at $6 a month. Prices change, so always check the current rates with ElevenLabs.

What v4 means for voice-overs in ads

For video ads, four things matter.

  1. Fewer pronunciation errors outside English. The biggest gain in our test: a voice that mispronounces a word costs trust.
  2. Emotion and pace live in the script. v4 has no style or speed sliders, and SSML (code such as break tags) is not supported, according to the ElevenLabs help center. So you write it into the text: [excited] for an energetic hook, [pause] right before your offer, [slowly] for a calm line and [quickly] to pick up the pace.
  3. Write a little shorter. Because v4 speaks more calmly, slightly less text fits into 15 seconds. Cut filler words or add [quickly] to lines that can move faster. A strong first line beats a lot of words. Our guide on how to make AI video ads shows how to build one.
  4. Know what v4 cannot do (yet). There is no v4 model for speech-to-speech, where you turn your own recording into another voice. For that, ElevenLabs still uses a v2 model. You can use your own voice through a voice clone on v4. Only clone voices you have permission to use.

Want to see how ElevenLabs compares with other voice tools for Dutch? Read Dutch AI voice over: 6 tools compared.

Frequently asked questions

Which languages does ElevenLabs v4 support?

More than 90, including English, Dutch, German and French. Both Eleven v4 and v4 Turbo share the same language list. You can pick from over 17,500 voices in the ElevenLabs library or use your own voice clone.

What is the difference between Eleven v4 and Eleven v4 Turbo?

v4 is built for the best quality in recorded audio such as voice-overs. v4 Turbo is faster, with about 150 milliseconds to first speech, and is meant for live conversations like a phone agent. On the API, Turbo costs half as much. For ads, choose v4.

Is ElevenLabs v4 better than v3?

In our own test with Dutch scripts, yes: 8 out of 8 takes correct on v4 against 3 out of 4 on v3. v4 also handles longer texts in one go and supports Professional Voice Clones again. The API price is the same.

Can I set the speaking speed in v4?

Not with a slider. v4 has no speed or style settings. You control pace with audio tags such as [slowly] and [quickly] in your text. Stability and similarity can still be adjusted.

The bottom line

Don't want to deal with API keys and per-character pricing? In Genly, personas and voice-overs have run on ElevenLabs v4 since September 29, 2026. You pick a script, a voice or an AI persona and get an on-brand video ad. You can start with a trial that includes 100 credits. Plans start at €39 a month; see the pricing page.

Share this article Copied
Marlon Bonink

Editor

Marlon Bonink

Co-founder of Genly

This article was written by Genly's AI team and reviewed by Marlon before it went live. How we work with AI: AI transparency.

Make ads like these for your business

Paste your website. Genly reads your brand and makes your first ads. Free to try.

Advertentie: Je ochtend, maar dan beter. Advertentie: Nieuw: Tropical Elixir. Advertentie: Zon in een fles.
Made with Genly for the demo brand Tropical Elixir

Read next