Leokani-tts 4.5M, and a second voice
Leokani's voice in a smaller model, and Pono, a male voice.
Two additions to Leokani-tts today.
Leokani-tts 4.5M is Dala, the Leokani voice, in 4.3 million parameters instead of 7. On one Arm core it speaks 38 times faster than real time, against 27 for the 7M model, and it reaches real time on an ESP32-P4. The words come out the same: 0.07 % word error rate, like the 7M.
Pono is a second voice for Leokani-tts 7M: male, American English, the same size and speed as Dala.
Both are free for non-commercial use on Hugging Face.