ElevenLabs

ElevenLabs turns text into speech that sounds like a person reading it, not a machine. It also clones a voice from a sample. There is a web app and an API, and the API is where the value is for me.

How I use it

Voice-over for video variants. A script of thirty seconds exists as text; I need the same script read in four tones, or four different openings on the same body, or the same video in German and English. A script sends each text to the API, gets back an audio file, and the video tool stitches it in. What used to be a studio booking per variant is now a loop. The multilingual voices are the quiet win: one voice character, several languages, and the German does not sound like a tourist.

What surprised me is the first time a batch of videos went out and nobody noticed the narration was synthetic. That was the point at which I stopped treating it as a novelty. What annoys me is the quota. Plans are measured in characters per month, and generating variants in bulk eats characters faster than you plan for, so I now count characters before a batch. Pronunciation is the other chore: a brand name, a product code, a price with a decimal, all need either a phonetic spelling or a pronunciation rule, and you discover each one by listening. The tiers also jump; the step from the Creator plan to Pro is a big one with little in between.

The web app is fine for a one-off. For anything repeated, use the API and keep the voice IDs in a config file.

Who it is for

Anyone making spoken video or audio in volume: ad variants, explainers, course material, a podcast intro you re-record often. Also anyone who needs the same narration in several languages. Skip it if you make one video a month; the free tier covers that, and so does your own voice. And skip it for anything where a listener should know it is synthetic; it is good enough that they will not.