ElevenLabs voice AI: what it does and what a minute of speech costs

ElevenLabs sells synthetic speech that listeners stop flagging as synthetic. Piotr Dąbkowski and Mati Staniszewski founded the company in 2022, school friends from Poland who had worked at Google and Palantir. The motive was personal: both grew up on films dubbed by one voice reading every part, and wanted dubbing to stop sounding like that.

“Voice-over people do not separate from a studio take”

— Intokened, Our card on the service

What it actually does

The product has three parts. It reads text in a chosen voice across more than 30 languages, holding the same voice through all of them. It clones a voice from a short recording, which is the feature people arrive for and the one carrying the consent problem. And it runs a separate model for dialogue, because two characters talking need timing a single reader never has to produce. Under all three sits one multimodal AI stack trained on speech rather than text alone.

Around that sit video dubbing, reading documents and books aloud, and an API, which is the part a developer needs. The service card in our catalogue lists the features and the checked prices in one place.

What a minute costs

Five tiers hold the pricing, and the number worth extracting is the price per thousand characters. Free gives 10,000 characters a month and asks for credit. Starter costs $6 for 30,000, which is $0.20 per thousand. Creator costs $22 for 100,000, or $0.22, and the company recommends it. Pro costs $99 for 500,000, $0.198. Scale costs $330 for 2 million, $0.165.

The discount for volume stays thin. Moving from Creator to Scale multiplies the allowance by 20 and cuts the unit price by 25%. Creator costs 10% more per character than Starter, so that step buys the professional clone rather than cheaper speech.

A thousand characters runs close to a minute of speech. That makes the free tier ten minutes a month, Creator about an hour and a half, Scale around 33 hours. A weekly podcast fits inside Creator, and a dubbing studio runs into Scale by the second week.

What to check first

Check three things before paying. The free tier asks for attribution and keeps commercial use out. Voice cloning starts at the $6 tier, not the free one. And the prices above come from the official page on 12 October 2026, which is why every card in our voice selection carries the date somebody looked.

Informational material, not a purchase recommendation. Plan prices come from the company's own pricing page and were checked on 12 October 2026; this market changes prices without announcements.