Make an AI voice greeting, read in one of eight voices

A message they play on speaker

A written greeting gets skimmed between notifications. A voice message gets played, often on speaker. Grandpa hears it without hunting for his reading glasses, a six-year-old hears it before she can read, and a friend plays it on the drive to work without taking a hand off the wheel.

The catalogue has eight voices made with ElevenLabs Eleven v3: Ava, Noa, Daniel, Mira, Tomer, Lena, Ari and Yael. Each has a short sample, so you choose by ear rather than by name, and the three above read the same line, so you can compare them one after another.

The price follows the length. A voice greeting is priced by the number of characters, so a two-line birthday wish costs little and a long retirement speech costs more. The text is free, the effects add nothing, the price is on the button before you press, and a recording that fails gives the credits back.

Daniel for the office card, Lena for a six-year-old

Eleven v3, from ElevenLabs, works on every plan and has the settings: eight voices, an accent for English and a Stability slider. A message six colleagues wrote together sounds odd read in one colleague's voice, while a neutral voice such as Daniel makes it from all of you. A six-year-old often listens better to a bright voice like Lena or Noa. The same eight voices speak in talking photo cards too.

An honest note: if you can record yourself today and like how it sounds, a voice memo from your phone is free and really you. A voice greeting here earns its place when the words need writing, or when the greeting has to land at breakfast in their city while you are asleep after a night shift or somewhere over the sea on a plane.

Name first, short sentences: writing for the ear

A greeting that reads well can sound stiff out loud. Short sentences help: the voice breathes at commas and full stops, so a line that runs for three rows comes out as one long breath. Put the name first - "Happy birthday, Rosa" - so she knows it is hers from the opening words. Then one real detail - the cherry tree she planted - one wish and a short close.

Write the words the way you want them said. "Seventieth" leaves no doubt where "70th" might, and a nickname spelled the way the family says it is the one to use. For a birthday, two to five sentences - about half a minute - is plenty; the agent keeps its spoken drafts that short. A wedding toast or a retirement speech can run longer: in Standard the limit is 2,500 characters, a few minutes of speech. Pick the voice after the words: the samples let you hear which one suits them.

Whoever reads it, the words should still sound like you. Your mother knows you have never once said "May your special day be filled with joy". Use the names you actually use - if the family calls Grandpa Deda, write Deda - and keep your habits: the way you open, the joke nobody else gets. Then read the text aloud once yourself; if it trips you up, it will sound stiff in any voice too.

When it needs a face, a tune or a party

A voice greeting is words and a voice, with nothing to look at. If the face matters - Grandma saying it herself - make a talking photo card: the same voices, up to 15 seconds of speech, animated by Kling AI Avatar v2 (face and head) or OmniHuman 1.5 (the whole person). It costs more, because every second of speech is also animated.

If you want the words sung, the song generator uses Google Lyria 3.5: you approve the lyrics first, and the song comes back in English, Russian, Hebrew, Arabic or another language. If you want the sound of the party rather than one voice - glasses clinking, the family singing the end of Happy Birthday - make a video: Veo 3.1 and Kling 3.0 make that sound along with the picture, in one generation. And if the words will be read, not heard, a text greeting costs nothing.