A modern alternative to Festival, MaryTTS and Pico TTS
Festival, MaryTTS, Flite, SVOX Pico and Mimic 3 are the classic free, offline text-to-speech engines: they run on small machines, need no cloud and have been in Linux distributions, Android and hobby projects for years. Several have stopped moving: MaryTTS last released version 5.2.1 in May 2022, Mimic 3 version 0.2.4 in November 2022 and Flite version 2.2 in 2020, while Festival is still maintained.
If you want the same offline independence with a more natural, modern voice, this guide shows how the SuperClonica engine fits and where the old engines still win.
What the classic engines are
Festival is the speech synthesis framework from the University of Edinburgh, with many voices built over the years. Flite is its small, fast relative for embedded use. MaryTTS is a multilingual engine written in pure Java. SVOX Pico is the compact engine that shipped with Android; on Linux its pico2wave command speaks US and British English, German, Spanish, French and Italian. Mimic 3 was Mycroft’s neural engine, released under AGPL-3.0.
Where they still win
They are tiny and fast, run on very old or very small hardware, start instantly and are packaged in many distributions. For a screen reader at high speed or a microcontroller-class device, a classic engine can still be the right tool.
A neural voice offline
The free SuperClonica engine runs the Supertonic 3 model on an ordinary processor, offline, in 31 languages. Install it with python -m pip install superclonica-cpu and voice a file with one command: superclonica --text-file notes.txt --voice F2 --lang en --output notes.wav. The ten stock voices sound like people rather than synthesizers and may be used commercially, and about 190 catalog voices add narrators, announcers and characters.
Replacing them in scripts
Where a script called pico2wave, flite or a MaryTTS server, call the superclonica command instead, or keep the engine loaded in Python for many short lines. On a Raspberry Pi, a 64-bit OS and a virtual environment work; see our Raspberry Pi guide. For a Home Assistant or Node-RED setup, the dedicated guides show the wiring.
Which to choose
Keep a classic engine for the smallest devices, extreme speech rates or a distribution package with no Python. Choose SuperClonica when the voice quality matters, for videos, audiobooks, assistants and games, and you still want everything to run offline.