Kokoro 82M โ€” bf16

Kokoro 82M in bf16 format optimized for MetalRT GPU inference on Apple Silicon. Includes 28 voice styles.

Usage

Used by RCLI with the MetalRT engine for text-to-speech:

rcli setup          # select MetalRT or Both engines

Performance (Apple M3 Max)

Metric Value
Latency 265 ms
RTF 0.10x
Voices 28
Format bf16

License

Model weights: Apache 2.0 (Hexgrad) MetalRT engine: Proprietary (RunAnywhere, Inc.)

Contact

founder@runanywhere.ai | https://runanywhere.ai

Downloads last month
14
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Collection including runanywhere/kokoro_bf16