Text-to-speech used to mean one flat voice reading your script. ElevenLabs Voice Engine v3 is the company's clearest attempt yet to close that gap between synthetic and human speech, and it's worth understanding what actually shipped before you decide whether to pay for it.
What's new in v3
The headline feature is emotional control through audio tags. Instead of feeding the model plain text, you can mark up a line to whisper, sigh, or land with more weight, and the model performs it instead of just narrating it. ElevenLabs also added multi-voice dynamic dialogue, so a single generation can carry a back-and-forth conversation between different voices without stitching clips together yourself.
Language support is now past 70 languages, and the company says it's not just translating; the emotional tone is meant to carry across languages instead of flattening into a neutral read.
Pricing, in plain numbers
Voice Engine v3 text-to-speech runs $0.10 per 1,000 characters on the Multilingual v2/v3 models. If you'd rather think in subscriptions:
| Plan | Price | Included characters/month |
|---|---|---|
| Free | $0 | Limited trial characters |
| Starter | $6/mo | 60,000 |
| Creator | $22/mo | 220,000 |
| Pro | $99/mo | 990,000 |
For context, 60,000 characters is roughly a 45-60 minute audiobook chapter, so the Starter plan covers casual use like short videos or podcast intros without much strain. See the full breakdown on our [ElevenLabs tool page](https://questloops.com/tools/elevenlabs).
How to try it for free
ElevenLabs still runs a free tier with a limited character allowance each month, no credit card required to start. That's enough to test the emotional tags and multi-voice dialogue on a short script before deciding whether the Starter plan is worth $6 a month. If you burn through the free characters fast, resetting the trial by creating a new account is against ElevenLabs' terms of service, so don't rely on that as a real strategy.
Is it worth switching to?
If you're already on an older ElevenLabs plan, v3's emotional tagging is a genuine upgrade for anything performative: audiobooks, character voices, ads with a specific tone. If you just need a flat, clear narration voice for tutorials or documentation, the improvement matters less and you might not need to pay for the jump at all.
For a broader look at what shipped across AI tools this month, see our [roundup of July 2026 launches](https://questloops.com/blog/5-ai-launches-from-july-2026-actually-worth-trying-and-what-they-cost). If you're weighing ElevenLabs against a video-first voice workflow, [ChatCut](https://questloops.com/tools/chatcut) is worth a look too.
