Read a script in any voice
Paste up to a full chapter, pick a voice, and get an MP3 back. Inline audio tags let you mark a whisper, a laugh or a pause instead of re-recording the take.
Text → Speech
Your script,
read out loud
in 74 languages.
ElevenLabs reads written text in a voice you pick, with control over pacing, emotion and pronunciation. Paste a script, choose a voice, export an MP3. This page explains what the tool actually does, what each plan costs, and where the limits are.
Free plan: 10,000 credits a month, about 10 minutes of speech. No card required.
Every word on this line started as plain text — and a model turned it into a voice.
74
Languages on Eleven v3
~75 ms
Latency on Flash v2.5
10,000
Free credits per month
$6
Cheapest commercial plan
What you get when you sign in
Six tools share one account and one credit balance. Most people come for the first one and end up using three.
Thousands of ready voices
Browse by age, accent, and delivery style — narration, news read, conversational, character. Save the ones you like so a series keeps the same narrator.
Use your own voice
Instant cloning needs about a minute of clean audio. Professional cloning, from the Creator plan up, trains on a longer sample and holds up over long-form narration.
Move a video to a new language
Upload a clip and get the same delivery in another language, with timing kept close to the original so it still lines up with what is on screen.
Edit long projects in one place
Split a script into chapters, assign a different voice per speaker, fix one sentence without regenerating the whole file, then export the finished track.
Wire it into your own product
Official Python and Node SDKs, streaming output so playback starts before generation finishes, and word-level timestamps for captions.
Four steps to your first voice track
Elapsed time from opening the signup page. Nothing to install.
00:00 → 00:40
Create the account
Sign up with email or Google. The free plan starts immediately with 10,000 credits — enough to test a voice properly before paying anything.
00:40 → 01:40
Pick a voice
Open the Voice Library and preview a few. Filter by language and by use case; a voice built for audiobooks will not sound right on a 30-second ad.
01:40 → 03:10
Paste the script and choose a model
Eleven v3 for expressive narration, Multilingual v2 for steady long-form, Flash v2.5 when speed matters. Add audio tags where you want a pause or a change in delivery.
03:10 → 04:00
Generate and export
Listen back, regenerate any line that lands wrong, then download the MP3 and drop it into Premiere, CapCut, Audition or your video editor of choice.
≈ 4 minutes · 0 dollars · 1 finished audio file
Which model to use
The model you select changes language coverage, speed, and how many credits each character costs. This is the part most new users get wrong.
| Model | Languages | Best for | Trade-off |
|---|---|---|---|
| Eleven v3 | 74 broadest coverage | Documentary narration, audiobooks, character dialogue, anything that needs real emotional range. | Slower to generate than Flash; overkill for short real-time replies. |
| Multilingual v2 | 29 stable set | Long-form voiceover where consistency across an hour matters more than expressiveness. | No audio tag control; flatter delivery on dramatic scripts. |
| Flash v2.5 | 32 ~75 ms latency | Voice agents, live apps, phone systems, and cheap drafts while you iterate on a script. | Lower expressive range — fine for information, thin for storytelling. |
Vietnamese is supported. A practical workflow: draft on Flash while you edit the wording, then run the final take on Eleven v3.
What it costs
Everything runs on one credit balance. On Multilingual v2, one character costs roughly one credit, and about 1,000 characters produces about a minute of speech.
| Plan | Per month | Credits | What it unlocks |
|---|---|---|---|
| Free | $0 | 10,000 ≈ 10 minutes | Full text to speech for testing. No commercial licence — you cannot monetise what you publish. |
| Starter | $6 | 30,000 ≈ 30 minutes | Commercial licence and instant voice cloning. The cheapest tier you can legally publish from. |
| Creator | $22 $11 first month | 121,000 ≈ 121 minutes | Professional voice cloning and higher quality audio output. The tier most solo creators settle on. |
| Pro | $99 | 600,000 ≈ 600 minutes | Volume tier for daily publishing, agencies, and heavier API use. |
Scale ($299) and Business ($990) sit above Pro; Enterprise is quoted directly. Annual billing bills ten months instead of twelve.
Figures taken from elevenlabs.io in August 2026 — check the pricing page before you buy, since tiers change.
Who this is actually for
The people who get the most out of it publish on a schedule and cannot record every take themselves.
YouTube & short-form
One narrator voice across an entire channel, including days you have a sore throat.
Audiobooks
Chapter-by-chapter production in Studio, with the same voice holding for ten hours of audio.
E-learning
Course modules that get re-recorded whenever the slide text changes — without booking a studio.
Ads and promos
Five versions of a 30-second read in five voices, tested before anyone commits budget.
Localisation
Publish the same video in Vietnamese, English and Japanese from one source script.
Voice agents
Support lines and in-app assistants that answer fast enough to feel like a conversation.
Questions people ask first
Is there really a free plan?
Yes, and it does not expire. You get 10,000 credits every month, roughly ten minutes of speech, with no card required. The catch is licensing: free-plan audio carries no commercial rights, so it is for testing rather than publishing. Commercial use starts on the $6 Starter plan.
Does it speak Vietnamese well?
Vietnamese is supported and the newer models handle it far better than earlier ones. Quality still depends on the voice you choose — a voice trained mainly on English speech carries an English accent into other languages. Preview several Vietnamese-native voices before committing to one for a series.
Can I use my own voice?
Instant voice cloning is available from the Starter plan and needs about a minute of clean recording. Professional voice cloning starts on Creator, trains on a longer sample, and is the one to use if you narrate long-form content. You may only clone a voice you own or have written permission to use.
What happens when I run out of credits?
On Free and Starter, generation simply stops until your plan renews. From Creator upward you can switch on usage-based billing and keep going, paying overage at a rate that depends on the model. Credits are shared across every tool, so dubbing and sound effects draw from the same balance as text to speech.
Can I monetise videos that use these voices?
Yes, on any paid plan — the commercial licence covers the audio you generate. Platform rules are separate: YouTube, for example, expects synthetic voices to be disclosed in some contexts, so check the policy for wherever you publish.
Do I need to know how to code?
No. Everything described on this page works in the browser. The API exists for people building the tool into a product, and it is optional.
Ten free minutes is enough to know.
Sign up, paste one paragraph you have already written, and listen. If the voice is not right for your work, you will know inside five minutes and it will have cost you nothing.
Create a free ElevenLabs account → No card required on the free planAffiliate disclosure. Nguyễn Luân ICT participates in the ElevenLabs affiliate programme. If you create an account through a link on this page, we may earn a commission at no additional cost to you. This page is an independent guide and is not written, endorsed or reviewed by ElevenLabs. Prices, credit allowances and language counts are quoted from elevenlabs.io as of August 2026 and can change — always confirm on the official site before purchasing. ElevenLabs is a trademark of its respective owner.