What is Eleven v3?
Eleven v3 turns text into speech that acts: it follows emotional cues in the script and square-bracket audio tags such as [whispers] or [laughs], and its Text to Dialogue API voices several characters in one generation. ElevenLabs now lists it as its previous generation, replaced by Eleven v4, and it costs $0.08 per 1,000 characters on the API, the same as Multilingual v2.
The trade-off is control and consistency. Each request stops at 5,000 characters, results depend heavily on the voice you pick, and for live voice agents ElevenLabs offers a faster sibling, v3 Conversational, at half the price.
How is Eleven v3 different from Multilingual v2?
Eleven v3 is more expressive and speaks far more languages; Multilingual v2 is steadier on long narration and takes longer requests, at the same price.
- Languages
- 74 for v3, against 29 for Multilingual v2.
- Request length
- Up to 5,000 characters per request on v3, against 10,000 on v2.
- Direction
- v3 takes audio tags for emotion and pacing; v2 uses SSML break tags for pauses instead.
- Dialogue
- v3 can voice several speakers in one generation through the Text to Dialogue API.
- Consistency
- ElevenLabs describes v2 as stable on long-form audio, and v3's most stable setting behaves most like it.
- Price
- Both cost $0.08 per 1,000 characters on the API.
Who is Eleven v3 best for?
Eleven v3 is best for audiobooks, games and video voiceovers that need acted, emotional delivery in many languages, rather than the fastest or cheapest speech.
Audiobooks and long narration
It follows emotional cues in the text and audio tags such as [softly] or [laughs warmly], so a narrator can change tone within a scene.
Games and character dialogue
The Text to Dialogue API voices several characters in one generation, and ElevenLabs names character scenes as one of the model's main uses.
Voiceovers in many languages
It speaks 74 languages, against 29 for Multilingual v2 at the same price per character, with more than 10,000 library voices to pick from.
Expressive voice agents
The v3 Conversational version answers in about 280 milliseconds at half the price per character, for support agents and interactive characters.
What's included with Eleven v3?
Audio tags for emotion and pacing, multi-speaker dialogue, three stability modes, IPA pronunciation and one-click tag suggestions in the app.
- Audio tags
- Tags in square brackets, such as [whispers], [sarcastic] or [laughs], set emotion, delivery and pace inside the script.
- Text to Dialogue API
- Voices a scene with several speakers in one generation, each speaker assigned a voice from the library.
- Stability modes
- Three settings, from Creative to the most stable, trade expressiveness against consistency; the most stable behaves most like v2.
- IPA pronunciation
- Write a word in IPA between slashes to fix how it is said, which ElevenLabs puts at 80 to 90% consistent.
- Tag suggestions in the app
- The Enhance button uses an LLM to add fitting audio tags to your text before you generate.
How much does Eleven v3 cost?
Eleven v3 costs $0.08 per 1,000 characters on the ElevenLabs API, about $0.08 a minute of speech; the free plan includes 10,000 characters a month, and paid plans start at $6 a month.
- 10,000 characters of v3 a month, about 10 minutes
- No commercial license
- Pay as you go top-ups from $5
- About $0.08 a minute of speech
- Up to 5,000 characters, about 5 minutes, per request
- v3 Conversational: $0.04 per 1,000 characters
- 75,000 characters of v3 a month on the API, about 75 minutes
- Commercial license
- Instant voice cloning
- 275,000 characters of v3 a month on the API, about 275 minutes
- Professional voice cloning
Pro is $99 a month (1,238,000 v3 characters on the API), Scale $299 (3,738,000) and Business $990 (12,375,000); Enterprise is priced on request. Yearly billing gives two months free, so Starter works out to $5 a month, Creator $18.33 and Pro $82.50. Every plan includes twice as many characters of v3 Conversational. In the ElevenLabs app the same plans come as credits instead (10,000 on Free, 30,000 on Starter, 121,000 on Creator, 600,000 on Pro), about 10, 30, 121 and 600 minutes of speech. Startups can apply for a grant of 33 million characters over 12 months. Eleven v3 is a closed model, reached only through the ElevenLabs app and API. Prices are in US dollars and exclude tax. Full pricing on elevenlabs.io. We checked these prices on .
How do readers rate Eleven v3?
No one has rated Eleven v3 yet. Be the first.
Sign in to rate. One rating per account.
Eleven v3 pros and cons
Listeners rate its English delivery close to human actors, but it can overact, drift between takes, sound accented with the wrong voice, and burn credits on retries.
Pros
- Listeners on Hacker News and Reddit rate its English delivery close to professional voice actors, and many call it the most expressive text-to-speech they have heard.
- Audio tags and multi-speaker dialogue let one generation act out a scene, which users point to for game characters and audiobooks with a voice per character.
- Testers praised some non-English output, such as Polish and Tamil, and one found Romanian worked well after switching to a native Romanian voice.
Cons
- Delivery can sound overacted: listeners describe forced laughs and an ad-read tone, and the voice can drift between takes or away from a cloned voice.
- Non-English speech often carries an English accent unless you pick a voice native to that language; testers heard it in French, Russian, Italian, German and more.
- Users call ElevenLabs expensive next to OpenAI and open models, and several say they burn credits regenerating lines until one sounds right.
- Each request stops at 5,000 characters, half of Multilingual v2, SSML break tags do not work, and ElevenLabs says professional voice clones are not yet fully tuned for v3.
What are the best Eleven v3 alternatives?
Eleven v4 for new projects at the same price, Sonic 3.6 for more minutes per dollar, and Voxtral TTS for a fifth of the price per character.
Sonic 3.6FreeChoose it when you want more minutes for the money; Cartesia's $5 Pro plan includes about 133 minutes of speech a month with instant voice cloning, against about 75 minutes of v3 on the $6 Starter plan.
Voxtral TTSFreeChoose it when price per character matters most and your languages are covered; Mistral charges $0.016 per 1,000 characters, a fifth of v3's rate, with voice cloning in 9 languages.
Eleven v4FreeChoose Eleven v4 for a new project; it replaces v3 at the same $0.08 per 1,000 characters, takes twice as much text per request and covers 90+ languages.
Questions about Eleven v3
What is the difference between Eleven v3 and Eleven v3 Conversational?
Eleven v3 is built for recorded work such as audiobooks, character scenes and voiceovers, and costs $0.08 per 1,000 characters on the API. Eleven v3 Conversational is the low-latency version for live voice agents, answering in about 280 milliseconds at $0.04 per 1,000 characters.
Who is Eleven v3 for?
Eleven v3 is best for audiobooks, games and video voiceovers that need acted, emotional delivery in many languages, rather than the fastest or cheapest speech.
Is Eleven v3 free?
Yes. Eleven v3 has a free plan: 10,000 characters of v3 a month, about 10 minutes.
How much does Eleven v3 cost?
Eleven v3's paid plans are API, pay as you go at $0.08 per 1,000 characters, Starter at $6 per month and Creator at $22 per month.
What are the best Eleven v3 alternatives?
The closest alternatives on Recaply are Sonic 3.6, Voxtral TTS and Eleven v4.
Where can I try Eleven v3?
On its official website, elevenlabs.io.
Can I use Eleven v3 audio commercially?
Only on a paid plan. ElevenLabs' Free plan comes without a commercial license; the Starter plan at $6 a month adds one, along with instant voice cloning, and every higher plan keeps it.
Why doesn't Eleven v3 sound like my voice clone?
ElevenLabs says professional voice clones are not yet fully optimized for v3 and can come out lower in quality than on older models. It suggests an instant voice clone recorded with a wide emotional range, or a designed voice, and a more stable setting on the stability slider for steadier output.
Does Eleven v3 support SSML?
Not for pauses: v3 ignores SSML break tags, so you set pacing with audio tags, ellipses and sentence structure. For pronunciation it reads IPA written between slashes instead of phoneme tags, which ElevenLabs puts at 80 to 90% consistent.
Our verdict
Eleven v3 is ElevenLabs' previous expressive model, still the pick when a script needs acting and your voices are tuned for it.
Listeners rate its English delivery close to human actors, it covers 74 languages, and it now costs the same $0.08 per 1,000 characters on the API as Multilingual v2.
- Skip it for
- New projects, which should start on Eleven v4 at the same price, live voice agents, and long narration where one steady voice matters more than range.
- Check first
- Test your exact voice on the free plan first; clones made for v3 must be retrained before they work on v4.
ElevenLabs Voice Design V3
Eleven Realtime
Eleven v4 Turbo
Scribe v2 Medical
Scribe v2
Fish Audio S2 Pro