Free · no sign-up · no credits

A free ElevenLabs alternative with no credits to run out.

For plain text to speech: paste a script, pick one of 28 English voices, and download a WAV or an MP3. No account, no monthly allowance, no watermark. The voice model runs in your own browser, so your text is never uploaded. It does less than ElevenLabs, and this page says exactly what.

0 / 100,000 characters

Voice

Speed

1.00x

Now hear it in any RVC voice.

That was one of our stock voices. In Xport Studio the same line can be read by any RVC model in your Library, including a clone of your own voice. No Python, no upload, offline on your Mac or PC.

Hear RVC voices read text Free to download · Pro is one time, never a subscription

The first run downloads the voice model once and your browser caches it: about 90 MB on most machines, about 310 MB on machines that run it on the graphics chip (WebGPU). After that it works offline with any voice you have already used. Up to 100,000 characters per run, English only, as many runs as you like.

What this is, and what it is not

People look for a free ElevenLabs alternative for one of three reasons: the monthly credits ran out, they do not want to make an account to hear one paragraph, or they do not want to send a private script to someone else's server. This page is built for those three cases. It is a text to speech tool and nothing else. You type, it speaks, you keep the file.

It is not a clone of ElevenLabs. There is no voice cloning, no dubbing, no sound effects, no API, and it reads English only. The voices are 28 stock voices from Kokoro, a small open source speech model. We have not run a listening test against ElevenLabs and we make no claim that these voices match it. What we can say is what the tool does, because you can check all of it yourself in the box above in under a minute.

ElevenLabs Free and Starter next to this page

The ElevenLabs columns below contain only what its own pricing page lists. Where that page does not state something, the table says so instead of guessing.

This pageElevenLabs FreeElevenLabs Starter
Price$0$0 per month$6 per month ($1 for the first month)
AllowanceNo credits. Up to 100,000 characters per run, any number of runs10k credits per month, about 10 minutes included30k credits per month, about 30 minutes included
Commercial useYes. Kokoro voices are Apache 2.0 and we add no restrictionA Commercial License is not listed on the Free plan cardCommercial License listed
Voice cloningNoNot listed on the Free plan cardInstant Voice Cloning listed
Beyond text to speechNothing. Text to speech onlySpeech to Text, Sound Effects, Voice Design, Music and more are listedEverything in Free, plus Dubbing Studio and more

ElevenLabs figures checked on 2026-11-11 against elevenlabs.io/pricing. Plans change, so confirm there before you decide. ElevenLabs is a trademark of its owner. Xport Studio is not affiliated with or endorsed by ElevenLabs.

When ElevenLabs is the better choice

If you need to clone a voice, dub a video, generate sound effects, or call an API from your own product, pay for ElevenLabs or a service like it. Its Starter plan lists a Commercial License, Instant Voice Cloning and Dubbing Studio, and none of those exist here. If you need a language other than English, this page cannot help you either. A free tool that pretends to cover those jobs would waste your afternoon.

When this page is enough

A lot of text to speech work is plainer than that. A scratch voiceover so you can time a video edit before recording the real one. A read-through of your own draft, to hear the clumsy sentences. Narration for a tutorial, a demo, or a slide deck. A placeholder vocal line while you write a hook. A long article you would rather listen to on a walk. For these, a monthly credit meter is the wrong shape of limit: you either do not use it, or you run out halfway through the one long script that mattered.

Here the limit is your own machine. One run takes up to 100,000 characters. The page splits the text at sentence ends, reads the pieces in order, joins them into one file, and shows the time left while it works. Cancel stops after the current piece and keeps what is done. Then you download a WAV for editing or an MP3 for sending, both made on your device.

How fast it is, measured

One real measurement, not a promise: on an Apple M4 Max laptop in a Chromium browser with WebGPU, a 6,201 character script came back as 6.6 minutes of audio in about 25 seconds, and encoding that as an MP3 took about one more second. A slower or older machine will take longer. A browser without WebGPU falls back to WebAssembly, which is slower than real time: with WebGPU switched off, the same laptop needed about 105 seconds to make 37.5 seconds of audio. The line under the tool tells you which one you are on; on WebAssembly, keep scripts short or expect a long wait. The first run also has to download the voice model: about 90 MB on most machines, about 310 MB on machines that run the full-precision model on the graphics chip. Your browser keeps it for next time.

Nothing is uploaded

A cloud text to speech service has to receive your words to speak them. This page does not, because the model is on your side of the connection. No request carries your text anywhere, and there is no account to tie it to. Once the model is downloaded, the tool keeps working with the network off for any voice you have already used (each voice is a small file fetched the first time you pick it). That matters for unreleased lyrics, client scripts, and anything under an agreement that says the text stays with you.

If you want a specific voice

This page only has stock voices. Xport Studio, our desktop app, reads text in any RVC voice model you own, including a clone of your own voice, on your own computer. It is free to download, and Pro is $79 once, not a subscription.

Related free tools

The same tool, with a longer guide to the voices, lives at free text to speech. To turn the WAV into M4A, FLAC, or OGG, use the audio converter. To trim a take, use the audio cutter, and to clean up a real recording, the noise reducer. Everything on the free tools page runs in the browser the same way.

Questions

Is this really a free alternative to ElevenLabs?

For plain English text to speech, yes. There is no account, no credit allowance and no watermark, and you download a WAV or MP3. It is not a replacement for everything ElevenLabs does: there is no voice cloning, no dubbing, no sound effects and no API here.

Is there a character or credit limit?

There are no credits. One run takes up to 100,000 characters, and you can do as many runs as you like. That cap is a memory safety limit on your device, not a quota. A long script takes a while, much longer without WebGPU, and Cancel keeps the audio made so far.

Does it sound as good as ElevenLabs?

We do not claim that. We have not run a listening test against ElevenLabs, and its paid product does far more than this page. These are 28 stock voices from Kokoro, an open 82M-parameter model. Generate a paragraph and judge with your own ears; it costs nothing to try.

Can I use the audio commercially?

Yes. The voices come from Kokoro, which is released under the Apache 2.0 licence, and we add no restriction of our own. There is no watermark and no attribution requirement on the audio you make here.

Can it clone my voice or a celebrity voice?

No, deliberately. This page only has stock voices. To read text in a voice model you own, including a clone of your own voice, use the Xport Studio desktop app, which does it on your own machine.

Is my text uploaded?

No. The voice model is downloaded to your browser once, and after that your text is turned into speech on your own device. No server receives your script, and the tool keeps working offline once the model is cached, with any voice you have already used.