Blog · AIKIZI Sound · ElevenLabs
Eleven v4: voices that act, not read
Write expressive Eleven v4 speech with audio tags, choose a voice and make your first clip in AIKIZI Studio. 10 free audio minutes daily this season.
Write the line the way an actor would need it: short sentences, real punctuation, and a bracketed tag where the feeling changes. Eleven v4 reads the tag as direction, not as text. Pick a voice that already sits near the character, then let the words do the rest.
Most text-to-speech reads. It gets the words right and the feeling flat. Eleven v4 was built for the other job: delivery. It hears a sigh in a sentence, it knows a question from a statement, and it will whisper if you ask it to. That makes it the first voice model where the script matters as much as the voice.
ElevenLabs is the company behind it, and for voice they are the reference the rest of the field measures against. AIKIZI's Studio runs v4 directly, with a curated set of voices and a switch that puts the audio tags at your cursor.
Why v4 is different
It performs the punctuation
A full stop lands. An ellipsis hangs. A question rises. Capital letters get weight. You are writing a performance, and v4 reads it as one.
It takes bracketed notes
Tags such as [whispers], [laughs], [sighs] or [excited] change how the next words are spoken. They are stage directions, and they are never read aloud.
It is multilingual
Write in the language you want spoken. The voice keeps its character across languages, which matters for a film with two tongues in it.
It holds a long take
Up to 10,000 characters in one request; Studio uses 5,000, about five and a half minutes of speech, so one take is one scene.
Audio tags: the director's notes
A tag is a word or two in square brackets, placed right before the words it should shape. Keep them sparse. One tag every few sentences reads as a performance; one on every line reads as a machine trying too hard.
| Tag | What it does | Use it when |
|---|---|---|
| [whispers] | Drops to a whisper for the phrase | A secret, a bedside line, a threat |
| [laughs] · [chuckles] | A real laugh, sized to the word | Between sentences, not inside them |
| [sighs] | An exhale before the line | Resignation, relief, a decision made |
| [excited] · [sarcastic] · [crying] | Colours the delivery of what follows | When the words alone could be read two ways |
| [strong French accent] | An accent on the voice, held for the take | A character, sparingly |
| [sings] | A sung phrase | A lullaby line, a hummed refrain |
Tags are hints, and some are less consistent across voices than others. If a tag does not land, change the voice before you change the words. A warm narrator will not do a good villain laugh, however you tag it.
Writing for a voice
Say it out loud first
If you run out of breath, the model will too. Break the sentence where you would breathe. A comma is a breath; a full stop is a pause.
Write numbers and names as sounds
"2133" becomes "twenty-one thirty-three" if that is how it should be said. Spell an unusual name the way it sounds, not the way it is written.
Put the emotion in the words, then the tag
"I said no." with [quietly] beats "[angry] I said no." A tag confirms a feeling the sentence already carries. It cannot invent one.
One take, one scene
Keep a take to a scene or a paragraph. If the tone changes, start a new take. Continuity between takes comes from the voice, not from cramming.
[sighs] You came back. I told you not to.
[whispers] The door on the left. Don't look at the window.
Go. [quietly] Go now.
Doing it in AIKIZI Studio
Open Studio, pick ElevenLabs Voice (V4) as the model, and choose a voice from the picker. Paste the lines. Turn on Audio tags and Studio inserts a tag at the cursor when you click one. Press Create. The take plays in the Sonic Field player and lands in your Library with a download.
While any free time remains, the Create button reads FREE and a chip beside it shows what is left today. A 5,000-character take uses about 5:35 of the 10 minutes. Your final free take can run longer than the time remaining: if 2:59 is left and the output lasts 3:30, the pool reaches zero for the day and no credits are spent. Ten minutes a day is roughly nine thousand characters of speech, which is a short film's worth of dialogue, every day.
Give a line to a voice
Ten minutes of Eleven v4 a day, free for every signed-in account this season. No card.
Open StudioHow the free minutes work
