CPU + 4 GB RAM
FP32 · Kokoro runtime
It should load, but 8 GB leaves safer room for the service and concurrent audio work.
Hexgrad
Audio model
Kokoro 82M
A tiny, fast speech model with a broad set of ready-made voices.
Speech
Capabilities
What you can use it for.
The most useful reasons to choose this model, without making you read through its repository first.
Text to speechMultiple voicesFast generationLightweight model
01
Narration
02
Accessible interfaces
03
CPU speech generation
For providers
Want to serve this model?
The provider requirements live here, separate from the information you need to choose and use the model.
Starting pointCPU · 4 GB RAM
Conservative estimate
CPU + 8 GB RAM
FP32 · Kokoro runtime
The main checkpoint is only about 327 MB and the official usage does not require CUDA.
Source and license
Check the original model.
OpenMayhem links back to the source repository so you can inspect the model card, files, license, limitations, and creator guidance directly.
