An alternative to Bark and Tortoise TTS that runs on a CPU
Bark and Tortoise TTS showed how natural open-source speech could sound, and both are still downloaded every day. In production they are hard to live with: Bark improvises, Tortoise needs an NVIDIA graphics card and patience, and neither repository has been updated since 2024.
The free SuperClonica engine runs the Supertonic 3 model on an ordinary processor, reads the text as written and voices long texts quickly. This guide compares the three honestly, including where Bark or Tortoise is still the better pick.
Why people look for an alternative
Both are important research projects; the trouble starts with long texts, many lines and ordinary computers.
- Bark is a text-to-audio model that, in its authors’ words, can deviate in unexpected ways from the script.
- Bark needs about 12 GB of video memory, or 8 GB with its smaller models, and is much slower on a processor.
- Tortoise needs an NVIDIA graphics card on your own computer and was built for quality, not speed.
- Both repositories were last updated in 2024.
Compare the sound free
On the page of any voice, paste a sentence you have voiced with Bark or Tortoise, 30 to 100 characters, choose the language and press Voice it. In a few seconds you can compare, with no account.
What changes with SuperClonica
Predictable output matters most when hundreds of lines have to be checked by ear.
- Install it with one command: python -m pip install superclonica-cpu, or run the Docker image as an HTTP service.
- A short phrase takes well under a second on an ordinary processor; in our tests 0.43 s on four threads of a Ryzen 7 5800X.
- The text is read as written, with no improvised sounds, and the same seed gives the same take again.
- Long texts are split at natural pauses and joined into one file.
Presets, clips and trained voices
Bark offers more than 100 speaker presets but no cloning of your own voice. Tortoise copies a voice from reference clips at every generation, on the graphics card. SuperClonica has ten stock voices and more than 170 catalog voices, each with a free demo; for your own voice, train it once in the SuperClonica studio from a clean 5 to 10 second recording of a speaker who agreed, on a computer with an NVIDIA graphics card, or order it on the exchange. After that the voice runs on any processor.
When Bark or Tortoise is still the better pick
Choose Bark when you need laughter, sighs, music or sound effects inside the audio, or Chinese, which Supertonic 3 does not speak. Choose Tortoise when you want to copy a voice from a few clips right away and have an NVIDIA card for every generation. Choose SuperClonica for clean, predictable speech in 31 languages at volume on ordinary hardware.