50% off yearly plansLimited time — ends in 50% off yearly plans · ends in

ElevenLabs v4 vs v3: The Same Tagged Script, Side by Side

One script with [whispers], [laughs] and [sighs] on Eleven v3, v4 and v4 Turbo: v4 beat v3 in every blind round.

  • Best voice: Eleven v4 Turbo
  • Best value: Eleven v4 Turbo
  • 3 models, one script
Made with Eleven v4 TurboEleven v4 Turbo's take of the shared script: one run, no edits.

Which AI model is worth it for expressive voiceover?

Best voice
Eleven v4 Turbo
Best value
Eleven v4 Turbo

v3 is not broken: it performed every tag. But in three blind listening rounds it came last each time, slower and flatter, with a whisper that was more quiet talking. v4 and v4 Turbo were close; Turbo ranked first twice and v4 once, and Turbo costs half as much.

Move from v3 to v4 for anything with tags; ElevenLabs also allows 10,000 characters per v4 request against v3's 5,000 (AdAnt runs up to 8,000). Use v4 Turbo by default and keep full v4 for the take you will publish if Turbo misses a nuance. Pauses work as on v3: SSML <break> is not supported, so use punctuation or a tag.

One expressive voiceover script, 3 AI voice models

We ran this script once on each model, with no retries and no edits: the same script and the ElevenLabs voice Jessica on each, through OpenRouter, the provider AdAnt runs these models on, on November 15, 2026. Listen to the whisper (a real whisper, or just quieter?), the laugh before "And honestly", the sigh, and the pause at the dots.

[excited] Okay, you guys, I have to show you this. [whispers] It's the one my sister keeps stealing. [laughs] And honestly? I'm not even mad. [sighs] Okay... I'm a little mad.
Eleven v40.07 credits · $0.018 · ready in 3 s

A true breathy whisper, a chuckle that runs into the line, a clear sigh and a beat at the dots. Natural and lively.

Eleven v4 TurboBest voiceBest value0.04 credits · $0.010 · ready in 2 s

Every tag, with the biggest contrast between the whisper and the excited lines. Ranked first in two of three blind rounds, at half v4's price.

Eleven v30.07 credits · $0.018 · ready in 4 s

It does every tag too, but slower (12.8 s against 11.6 s) and flatter, and its whisper is more hushed talking than a whisper. Last in every blind round.

Price difference: the same expressive voiceover script

The same read cost 0.04 credits on Eleven v4 Turbo and 0.07 credits on Eleven v3: 1.8× the price for one script.

ModelThis readIn dollarsvs cheapestReady inVerdict
Eleven v4 Turbo0.04 credits$0.010Cheapest2 sBest voice · Best value
Eleven v40.07 credits$0.0181.8×3 s—
Eleven v30.07 credits$0.0181.8×4 s—

Credits are what each run costs on AdAnt. Dollars are at the Standard plan's 26¢ a credit ($39 for 150 credits); bigger plans and yearly billing pay less per credit. A voice is priced by the character, so a longer script costs more, and the try-it chat shows the price before you run.

  1. Write your script with [tags] right before the words they should change.
  2. Pick a model. v4 Turbo is the default here; v3 is there to compare.
  3. Press Create, then compare the take with the three runs on this page.

ElevenLabs v4 vs v3: The Same Tagged Script, Side by Side

ElevenLabs released Eleven v4 and v4 Turbo on September 28, 2026, after v3. ElevenLabs says v4 follows audio tags such as [whispers] and [laughs] more reliably than earlier models. We ran one tagged script on all three, with the same voice, to hear whether that holds up, and what each costs.

UGC ads

A creator-style read with a laugh and a whisper.

Characters

Reactions on cue for games, cartoons and dramas.

Moving off v3

Hear your v3 script on v4 before you switch.

Eleven v4

Four audio tags on one line, pricing, API and voice cloning.

Try it as a tool

Eleven v4 vs v3 FAQ

Tips for a better result
  • Put each tag right before the words it should change.
  • Combine tags for a mix: [whispering, playful], or [nervous][whispers] as in our run.
  • Use punctuation or a tag for pauses; like v3, v4 does not support SSML <break>.
Is Eleven v4 better than v3?

In our blind test, yes: v3 came last in all three rounds, slower and flatter, while v4 and v4 Turbo whispered, laughed and sighed more naturally. v3 still did every tag.

What changed from Eleven v3 to v4?

As of November 2026, ElevenLabs says v4 follows audio tags more reliably than v3. It allows 10,000 characters per request (v3: 5,000) and comes with a faster, half-price v4 Turbo. Neither v3 nor v4 supports SSML <break>.

Do v3 and v4 cost the same?

On OpenRouter, Eleven v4 and v3 both list at $0.08 per 1,000 characters, and v4 Turbo at $0.04. Our 175-character script cost 0.07, 0.07 and 0.04 AdAnt credits.

How did you test Eleven v4 against v3?

We sent one tagged script once each to Eleven v3, v4 and v4 Turbo through OpenRouter on November 15, 2026, with the ElevenLabs preset voice Jessica on each. There were no retries or edits, and we ranked the takes in three blind listening rounds.

Compare models for another use case

Every model and use case

Want the video, not just the voice?

AdAnt's agents write the script, pick the right model for each shot and voice, and make the full video in one chat.

Start with AdAnt