A gTTS alternative with natural voices, offline
gTTS is the library many Python tutorials start with: it sends text to the speech service of Google Translate and saves an MP3. It is easy and free, but it needs the internet, it is not an official API and can be rate-limited or break, and it gives you languages and accents rather than a choice of voices.
SuperClonica is a free engine that voices text on your own computer, offline, with ten stock voices, character voices from the catalog and voices trained from your own recordings. This page compares the two honestly.
Two different setups
One borrows an online service; the other is an engine you install and keep.
- gTTS: text goes to an online service; one voice per language and accent; unofficial, so requests can be limited or fail.
- SuperClonica: the model runs locally after a one-time download of about 385 MB; nothing is sent anywhere.
- SuperClonica voices speak 31 languages; Chinese is not supported yet, while gTTS can speak Chinese.
- SuperClonica gives you many voices: stock, character and your own.
When gTTS is fine
For a quick script with an internet connection, a language Supertonic 3 does not speak, or when the voice simply does not matter, gTTS is a reasonable choice, and SuperClonica does not replace its range of languages.
When SuperClonica fits better
The same text, voice and seed always give the same audio, so results are reproducible.
- Your program must work offline or on a closed network.
- Your texts are private and should not leave the machine.
- You need a choice of voices, a character voice or your own trained voice.
- You voice a lot of text and do not want to hit limits of an unofficial service.
Switch in a few lines
Install the engine with python -m pip install superclonica-cpu. Where your code creates a gTTS object and saves it, use the engine instead: after from superclonica import Superclonica and tts = Superclonica(), write audio, rate = tts.synthesize(text, voice="F1", lang="en"); tts.save("hello.wav", audio). For an MP3 like gTTS gives you, run the command-line engine with an output name ending in .mp3. Long texts are split at natural pauses and joined into one file. Before you switch, try a voice on its page with a line of 30 to 100 characters, free and without an account.
What it costs
The engine and the ten stock voices of the model are free, and several catalog voices are free too. Other voices are bought once and stay yours. There is no per-request fee and nothing to rate-limit you: once installed, it works without the internet.