Free unlimited text to speech without sign-up: run it on your own computer
Most online text-to-speech services are free only up to a point: a few thousand characters a month, a sign-up form, then a subscription. That is not greed but cost: every character is synthesized on their servers.
There is another way that has no such limit. Supertonic 3, an open text-to-speech model, runs on an ordinary computer, and the SuperClonica engine and free voices are free to download. This guide shows how to set it up in a few minutes and voice as much text as you want, without an account.
Why online text to speech has limits
A web service synthesizes your text on its own graphics cards and pays for every second of their work. Free plans therefore cap the number of characters, require an account to count them, and leave the rest to paid plans.
When the model runs on your computer, the work is done by your processor. There is nothing to count and nothing to charge for, so there is no limit.
What you need
An ordinary computer is enough.
- Windows, macOS or Linux with Python 3.10 or newer, or Docker.
- About 400 MB of disk space for the Supertonic 3 model, downloaded once.
- A voice: a single JSON file. Free ones download without an account.
- No graphics card: an NVIDIA GPU only makes it faster.
Step 1: download a free voice
Open the free voices page of the catalog, listen to the demos and press Download for free on the voice you like. No sign-up is needed. The same free voices are in our GitHub repository with audio samples.
Supertonic 3 also ships with ten preset voices, five male and five female; their JSON files work with the engine too.
Step 2: install the engine
Install the SuperClonica engine with one command: python -m pip install superclonica-cpu. For an NVIDIA graphics card use superclonica-gpu-cuda instead; pip installs CUDA and cuDNN itself. If you prefer containers, the superclonica/superclonica image on Docker Hub runs the same engine as an HTTP server.
Step 3: voice as much text as you want
Voice a text with one command: superclonica --text "Your text" --voice voice.json --lang en --output speech.wav. The model downloads on the first run; after that the engine works without internet.
Long texts are no problem: the engine splits them at natural pauses and joins the parts into one file. On a desktop 8-core processor we estimate it at 6.4–9.7 times faster than real time, so an hour of audio takes roughly 6 to 10 minutes.
What exactly is free
Nothing here has a trial period or a character counter.
- The SuperClonica engine: free, under the Apache 2.0 license.
- The Supertonic 3 model: free, under the OpenRAIL-M license, which also allows commercial use with its use restrictions.
- Free voices from the catalog and the preset voices of Supertonic 3.
- One phrase on the page of any voice, voiced online for free.
- Paid voices are bought once, with no subscription: the file stays yours.
Free online options and their limits
If you cannot install anything, the voices built into your operating system and browser read text for free and without limits, but they sound noticeably robotic. Online neural services sound natural, but their free plans are limited and ask you to sign up. A local neural model gives both: natural speech and no limits.