How to use AI Voice Generator
Pick a voice and a language
Choose a voice under Voice and the language of your script under Language. Language starts on the language of this page. It tells the voice how to read the text and does not translate it, so write the script in that language. To compare voices, click a Style preset: it fills Text to speak with a sample script and replaces whatever is in the box.
Set speed and expression
Speed sets the pace: Slow for dense explanations, Fast for short promos. Expression sets how much the read varies. Steady keeps it even for tutorials and support. Natural suits most voiceovers. Expressive gives ads and characters more range, but two runs of the same line can differ more.
Steer the delivery with audio tags
Put a cue in square brackets right before the words it should change, e.g. "[whispers] Don't tell anyone. [laughs] I'm joking." Tags like [excited], [sighs] or [sarcastic] work best with Natural or Expressive, and Steady follows them less. Tags count as characters, and punctuation still sets the pauses.
Generate, listen and fix
Press "Generate voice" and listen for rushed passages and misread names. Respell a name the way it sounds, split long sentences, then generate again. One run takes up to 3,000 characters, so split a longer script into sections and redo only the section that needs it.
What a real job costs here
Runs on this page cost 15 credits. At the best pack rate that is about $0.1596.
| Pack | Credits | Thousand characters with ElevenLabs v3 |
|---|---|---|
| $9 | 750 | 50 |
| $19 | 1,650 | 110 |
| $39 | 3,500 | 233 |
| $79 | 7,300 | 486 |
| $149 | 14,000 | 933 |
What this model is good at and where it stops
ElevenLabs v3: Voiceovers, ads and narration that need emotion and many languages. Expressive delivery steered by audio tags. Watch out for: needs more script work than earlier ElevenLabs models.
Choose the voice first
Voices differ more than settings do. Take the hardest line of your script, the one with a name, a number or a joke, and generate it in two or three voices. Keep the one that gets it right for the whole project, so every section sounds like the same speaker.
Speed and expression, not just the words
Speed changes the pace, from Slow to Fast. Expression changes how much the voice varies: Steady for tutorials and support content, Natural for most voiceovers, Expressive for ads and characters. More expression means more difference between runs, so generate a key line twice and keep the better take.
Audio tags direct the performance
ElevenLabs v3 reads cues in square brackets, such as [whispers], [excited], [sighs] or [laughs]. Put a tag right before the words it should change and use one per sentence at most. A script full of tags sounds like overacting.
Write for the ear
Short sentences, one idea each, real punctuation. Write names and numbers the way they should be said. If a line is hard to read aloud, it is hard to listen to.
Then put it on camera
Pair the audio with lip sync so a presenter says it, or drop it under generated footage as a voiceover.
Frequently asked questions
How is it priced?
15 credits per 1,000 characters, roughly a minute of speech. Spaces, punctuation and audio tags count as characters, and each started 1,000 counts in full: 1,200 characters cost 30 credits. The cost updates as you type.
Which voices and languages can I use?
A set of female and male stock voices, and 14 languages: English, Polish, German, Spanish, French, Italian, Portuguese, Dutch, Swedish, Japanese, Korean, Chinese, Indonesian and Hindi. Language starts on the language of this page and sets how the text is read. It does not translate.
What are audio tags?
Cues in square brackets, such as [whispers], [excited] or [sighs], placed before the words that should sound different. They steer the tone of the next phrase without a separate setting.
How long can one script be?
Up to 3,000 characters per run, about three minutes of speech, for 45 credits. Generate a longer text in sections and join them in an audio or video editor.
Can I clone my own voice?
Not on this page. It uses stock voices only.
Can I use the audio commercially?
Yes on paid packs, under the UGESI terms.
