What is ElevenLabs v3?
ElevenLabs v3 (Eleven v3 in ElevenLabs' own docs) is a text-to-speech model built for expressive reading. ElevenLabs describes it as "emotionally rich" with "high emotional range and contextual understanding across 70+ languages", and says it works well for character discussions, audiobook narration and emotional dialogue.
It's the generation before Eleven v4. ElevenLabs now points most people to v4, yet says the two can sound quite different and you may still prefer how a voice sounded in v3. Bolly runs both, so you can judge by ear: see ElevenLabs v4.
On Bolly, v3 works as a plain AI voice generator: your words go in, you choose a voice, and an MP3 comes out. Bolly is an independent studio with no tie to ElevenLabs.
The ElevenLabs v3 voices list
Bolly gives v3 a menu of 21 preset voices: pick Rachel, Aria, Roger, Sarah, Laura, Charlie, George, Callum, River, Liam, Charlotte, Alice, Matilda, Will, Jessica, Eric, Chris, Brian, Daniel, Lily or Bill. Rachel is selected when the page opens; the microphone chip on the toolbar switches it.
That menu is the whole selection: Bolly has no voice cloning and can't reach voices saved in an ElevenLabs account.
A name says little about a sound, so audition: run one paragraph from your project in two voices, which a day's free credits cover. Pick the voice before you write tags. ElevenLabs says tags depend on the voice and the context, and that a delivery already in a voice's training data, such as whispering or shouting, is easier to reproduce.
What the free daily credits cover
Sign in with Google and Bolly gives your account 10 free credits a day. ElevenLabs v3 is billed per block of 1,000 characters, at 4 credits a block, and everything in the box counts, spaces and tags included. A day covers:
- Two takes of up to 1,000 characters, 4 credits each.
- Or one take of up to 2,000 characters for 8 credits: about two minutes of speech, by ElevenLabs' estimate of five minutes per 5,000 characters.
- The 2 credits left over won't buy a third v3 take, but they cover 1,000 characters on Qwen Audio 3 TTS.
The 10 credits return at midnight UTC, and whatever you didn't spend lapses then. Longer projects call for monthly credits: the Starter plan adds 300 a month for $19, which covers 75 v3 takes of up to 1,000 characters. See all plans.
The Free plan also covers one free character swap video, in which Bolly casts you in a viral clip from one photo. It unlocks once 24 hours have passed since you joined. Viggle makes it, a Bolly watermark stays on it, and it spends 5 credits of the 10 you have that day, still enough left for one v3 take.
ElevenLabs v3 text to speech, step by step
Type what you want to hear
Paste or type your script above. Enter starts a run, so use Shift+Enter for line breaks. ElevenLabs takes up to 5,000 characters per v3 request; split anything longer.
Pick the voice
Click the voice chip, which reads Rachel at first, and choose one of the 21 names.
Open the extras if you need them
The small up-arrow beside the credit price opens Additional settings: Stability, Apply text normalization and Language. You can leave all three at their defaults.
Run it and save the MP3
Press the green arrow button at the far right. On your first run a sign-in box opens: continue with Google, and the take then plays under the text box. Download saves an MP3, and your Library keeps every take.
Audio tags: stage directions for v3
Audio tags are cues in square brackets that v3 performs instead of reading out. ElevenLabs groups them as emotions, such as [sad], [angry] and [happily]; delivery, such as [whispers] and [shouts]; and reactions, such as [laughs], [clears throat] and [sighs]. A tag shapes the words that follow it.
SSML break tags aren't supported in v3, so build pauses into the writing with ellipses, short sentences and new paragraphs. Each run speaks in one voice: for a scene with two characters, give each part its own run, then join the files in an audio editor.
[curious] Nobody knows who planted the first orchard on Wren Hill. The trees are older than the village, older than the road that leads to it. [sighs] Every autumn the apples fall, and every autumn someone new arrives to ask about them... [whispers] and most of them decide to stay.
[excited] Meet the Sunday Box from Tallowfen Bakery. Six loaves, still warm, a jar of wildflower honey and a handwritten recipe card, on your doorstep by eight. [whispers] The cinnamon swirl sells out by Thursday. [happily] Order by Friday noon. Tallowfen Bakery. Baked at four, with you by breakfast.
[clears throat] Right. You must be the new apprentice. [sarcastic] Wonderful. Another one. [sighs] Rule one: never feed the dragon after dark. Rule two: if it smiles at you, back away slowly... [shouts] And rule three. Close. The. Door.
Stability, language and numbers
- Stability: a number from 0 to 1 (default 0.5). ElevenLabs says lower values widen a voice's emotional range and higher ones can sound monotonous, so go down for drama and up for a steady read.
- Language: optional. Enter an ISO 639-1 code (two letters, like ta for Tamil or ko for Korean) to make v3 use that language. ElevenLabs lists 70+ languages for v3, Arabic, Ukrainian, Welsh and Yoruba among them.
- Apply text normalization: on auto, the model decides whether to spell out things like numbers; set to on, it always does, and set to off, it never does.
Put the voice to work
A downloaded take works in other Bolly tools. Create Video Avatars turns a photo and a voice track into a talking avatar, and Lipsync Video re-voices a video so the lips match. Both need the Starter plan or higher. The AI movie guide walks through a voice-over made with v3.
Voicing real people
Make a script speak for a real person only if they've agreed, and mark realistic audio as AI-generated wherever you share it. Read the Responsible AI policy for the full rules.
Frequently asked questions
Is ElevenLabs v3 free on Bolly?
Yes, within the Free plan's daily credits. Signing in with Google gives you 10 credits each day, and v3 uses 4 for each 1,000 characters. There's nothing to pay to start, and the allowance renews daily.
How much text can I turn into speech free each day?
Up to 2,000 characters, roughly two minutes of audio: two runs of up to 1,000 characters, or one of up to 2,000.
Are my free audio files private?
No. Free-plan takes are always stored as public, even if you flip the globe button in the toolbar, and they can show up in Explore. Private audio starts with Starter, the entry-level paid plan.
Which voices can I use?
The 21 voices in the v3 menu, from Rachel to Bill. You can't add your own, since Bolly has no voice cloning. To get a voice from a written description instead, try Voice Designer.
Can I download the MP3?
Yes. Once a take is ready, press Download under the player to save it as an MP3. Every take also stays in your Library to play or download again later.
Is this the official ElevenLabs site?
No. You're on Bolly, an independent AI studio that isn't affiliated with ElevenLabs. Bolly runs the v3 model and charges in its own credits. For an ElevenLabs subscription, voice cloning or API access, go to elevenlabs.io.
ElevenLabs v3 vs v4: what's the difference?
ElevenLabs calls v4 "a net upgrade over Eleven v3" and recommends it for most work. On Bolly, v4 adds a Turbo mode at 2 credits per 1,000 characters plus Output format, Similarity boost and Seed settings, and a 2,000-character script costs 7 credits on v4 versus 8 on v3. v3 keeps the voice menu; on v4 you type the name. Hearing one short script in both costs 8 credits, within a single free day.
How long can one ElevenLabs v3 script be?
ElevenLabs allows 5,000 characters per v3 request, about five minutes of speech. On Bolly a full-length run costs 20 credits, beyond a single day's free allowance, so the longest free run is 2,000 characters.




