To use ElevenLabs, create a free account, open the Text to Speech tool, paste in a script, pick a voice from the library, and hit Generate — the audio downloads as an MP3 in seconds. Everything past that first generation (cloning your own voice, dubbing a video, building a conversational agent) builds on that same core loop.
Short answer: Sign up free at elevenlabs.io, open Text to Speech, paste your script, choose a voice from the library (or clone your own), and click Generate. The free plan gives you 10,000 credits a month with no commercial license; $6/month Starter adds instant voice cloning and commercial rights. Most people get a usable clip on their first try.
In my testing, the fastest way to waste time with ElevenLabs is skipping straight to voice cloning before you've generated a single clip with a stock voice — the platform makes more sense once you've seen how stability and similarity sliders change a normal voice first. I ran the same 200-word script through five of the library's premade voices and two of the paid tiers to see where the free plan actually runs out, and what follows is the order I'd walk a first-time user through it.
What you'll need
An email address is enough to start — ElevenLabs' free plan needs no credit card and gives you 10,000 credits a month, roughly 10,000 characters of speech on the standard model. You don't need any audio equipment for basic text-to-speech, but if you want to clone your own voice later, a quiet room and a decent phone microphone will get you through instant cloning; professional cloning wants 30-plus minutes of clean recorded audio. Have your script ready before you sit down — pasting in a finished paragraph, rather than typing live, is where you'll actually judge how a voice sounds, since ElevenLabs' delivery changes with punctuation and sentence length.
Step-by-step: how to use ElevenLabs
1. Create a free account
Go to elevenlabs.io and sign up with an email or a Google account. No credit card is required for the free plan, and you land straight in the dashboard with Text to Speech, Speech to Text, Voice Cloning, Dubbing, Music, Sound Effects, and Conversational Agents all listed as separate tools down the left side.
2. Open Text to Speech and paste your script
Click Text to Speech, paste your script into the text box, and leave the model on its default for your first run. In my testing, punctuation matters more than most people expect — a script with proper commas and periods reads far more naturally than one long unbroken sentence, because the model uses that punctuation to decide where to breathe and pause.
3. Pick a voice from the library
Click the voice selector and browse the library — ElevenLabs lists over 10,000 voices, covering 70-plus languages and a wide range of English accents, according to ElevenLabs’ Text to Speech page. Preview a few before committing; a voice that reads a short greeting well doesn't always hold up over a full paragraph, so preview with a chunk of your actual script, not the sample line.
4. Tune stability, similarity, and style
Open the voice settings panel and adjust Stability (how consistent the delivery is take to take), Similarity (how closely it sticks to the reference voice), and Style Exaggeration (how much emotional inflection comes through). In my testing, pushing stability too low made a normally calm voice sound erratic on longer scripts, while pushing it too high flattened emotion out of dialogue-heavy text — the useful range for most narration sits in the middle of the slider, not at either end.
5. Generate and download the audio
Click Generate. A first pass on a short paragraph takes a few seconds; longer scripts take longer but still land well under a minute in my testing. Download the result as an MP3, or regenerate with tweaked settings if the delivery didn't land — regenerating only spends credits on the new attempt, not the ones before it.
6. Try voice cloning, dubbing, or agents once the basics click
Once Text to Speech feels routine, Instant Voice Cloning (a paid feature starting on the $6/month Starter plan) lets you clone a voice from a short sample in a couple of minutes; Professional Voice Cloning wants 30-plus minutes of audio for a more accurate result. Dubbing Studio translates and re-voices a video into another language automatically, and Conversational Agents let you build a voice bot that can hold a two-way conversation instead of just reading a script aloud.
Example prompts and scripts you can copy
These are close to what I actually pasted into Text to Speech while testing — adjust the wording, but keep the punctuation and audio tags, since that's what actually shapes delivery:
- Narration test: "Here's the thing nobody tells you about learning a new skill. [pause] It's not the hard days that stop people — it's the boring ones."
- Emotional range test: "[excited] I can't believe we actually pulled this off. [laughs] Six months of work, and it's finally live."
- Customer-facing script: "Thanks for calling. [warm] I can help you with that — can you give me your order number when you're ready?"
- Whisper/ASMR-style test: "[whispers] Close your eyes for a second. [sighs] Just breathe."
- Voice-cloning sample line (read this aloud to record your own clip): "The quick brown fox jumps over the lazy dog while the rain falls steadily outside the window."
Bracketed tags like [pause], [laughs], [whispers], and [excited] are inline audio directions ElevenLabs' models respond to — they don't always trigger identically across voices, so preview before you commit to a final script.
Common mistakes to avoid
The mistake I made most often early on: burning free-plan credits on full-length drafts instead of testing a 20-word sample first — at 10,000 credits a month, three or four false starts on a long script add up fast. Second, expecting instant voice cloning to sound exactly like the recorded sample on the first try; it's close, not identical, and professional cloning with a longer sample closes that gap if the difference matters for your project. Third, ignoring punctuation and treating the text box like a plain notepad — periods, commas, and ellipses are doing real work in how the model paces a line, and a wall of text with no punctuation reads flat no matter how good the voice is. Fourth, forgetting that the free plan has no commercial license — anything you plan to publish, sell, or monetize needs at least the $6/month Starter plan. Fifth, sticking with the default voice settings for every project instead of adjusting stability and style per script; a calm narration voice and an excited ad-read need different slider positions to sound right.
ElevenLabs pricing (2026)
| Plan | Price | Credits/month | Voice cloning | Commercial use |
|---|---|---|---|---|
| Free | $0 | 10,000 | Not included | No |
| Starter | $6/mo ($5/mo billed yearly) | 30,000 | Instant | Yes |
| Creator | $22/mo ($18.33/mo billed yearly) | 121,000 | Professional | Yes |
| Pro | $99/mo | 600,000 | Professional | Yes |
| Scale | $299/mo | 1,800,000 | Professional (3 clones), 3 seats | Yes |
Prices and credit allowances above are the individual plans listed on ElevenLabs’ pricing page as of July 2026; Business and Enterprise tiers exist above Scale for teams needing more seats and support, but most solo users won't need to go past Creator or Pro.
Tools that make this easier
ElevenLabs handles the voice side of content creation, but it's usually one piece of a bigger workflow — if you're scripting narration or ad copy before you record it, my best AI writing tools roundup and hands-on Jasper review cover what's worth pairing it with for the writing itself. If your voice work is one part of a larger content pipeline — blog posts, video scripts, social captions — best AI tool for content creation compares the options across formats. For drafting a script through conversation before you paste it into Text to Speech, how to use Claude AI is a solid option for long-form writing. Comparing ElevenLabs against a built-in voice assistant rather than a standalone generator is also worth doing: how to use ChatGPT Voice covers OpenAI's conversational voice mode, which solves a different problem (talking to an assistant) than ElevenLabs (generating a voiceover you control and keep). And if you're weighing ElevenLabs against other AI tools before committing to a paid plan, my AI tool reviews hub has hands-on tests across categories.
My take
ElevenLabs is the most natural-sounding text-to-speech I've tested, and the free plan is genuinely enough to figure out whether it fits your project before you pay anything. Where it earns real money is voice cloning and dubbing — Starter's $6/month unlocks the commercial license almost everyone actually needs the moment they publish anything, which makes it the plan most people should land on rather than staying on Free indefinitely. If you're doing high-volume work — a daily podcast, a YouTube channel with long scripts — Creator's 121,000 monthly credits is where the math starts working out better than paying overage rates on Starter.
Frequently Asked Questions
Is ElevenLabs free to use?
Yes, with limits. The free plan gives you 10,000 credits a month (roughly 10,000 characters of speech) and no credit card required, but it doesn't include a commercial license or voice cloning — those start on the $6/month Starter plan.
How long does it take to learn how to use ElevenLabs?
Generating your first clip takes a couple of minutes — sign up, paste text, pick a voice, generate. Getting a feel for the stability and style sliders, and how punctuation changes delivery, takes a handful of test scripts, usually under an hour of hands-on use.
What's the easiest way to get started with ElevenLabs?
Sign up free, paste a short paragraph of your actual script into Text to Speech, and try three or four different premade voices before touching any settings. Once you know which voice fits, then adjust stability and style to match the tone you want.
Can I clone my own voice with ElevenLabs?
Yes. Instant Voice Cloning, available from the $6/month Starter plan up, creates a clone from a short sample in a couple of minutes. Professional Voice Cloning, available from Creator and up, wants 30-plus minutes of clean recorded audio and produces a closer match.
Does ElevenLabs support languages other than English?
Yes. ElevenLabs supports speech generation in 70-plus languages, according to its Text to Speech product page, and its Dubbing Studio tool can translate and re-voice existing audio or video into another language automatically.