New Free AI Turns Text into Speech with Voice Cloning, No GPU Needed
Save money on voice AI and run it on your laptop—no expensive hardware required.
KittenML has released the code and weights for Kitten-TTS 2, described as a super-efficient frontier TTS model and the world's first 1.58-bit TTS model. According to the article, it runs in real time on just a CPU—no GPU needed for real-time generation—and on consumer GPUs too.
It uses ternary weights (-1, 0, +1) with group size of 128 for scales, has 1.7B parameters, and offers zero-shot voice cloning from less than 5 seconds of reference audio without fine-tuning. It supports voice cloning in 20+ languages, cross-language cloning for accented speech, and expressive/emphasis control to steer delivery and stress specific words. The code and weights are open-source and free to use, and it can be self-hosted or run locally for super-cheap use, positioned as an OSS alternative to ElevenL
- Runs on a regular CPU in real-time, so no expensive GPU needed—saves money on hardware.
- Clones any voice from just 5 seconds of audio and speaks 20+ languages with emotion control.
- Completely free and open-source, letting anyone self-host or build apps without paying for cloud services.
Why It Matters
Makes advanced voice cloning and multilingual speech free and accessible, enabling cheap local apps and privacy-friendly tools.