ZMIME

tiny-voca-arabic vs ZONOS3-Mini

Compare documented capabilities and access. A dash means a comparable figure is unavailable.

Model specifications
Specificationtiny-voca-arabicoddadmixZONOS3-MiniZyphra
TaskSpeech synthesisSpeech synthesis
Published capabilitiesA compact CPU-oriented model trained from scratch using the tiny-voca architecture and an Arabic phonetic front end.Supports promptable voice style and speech editing. Runs locally, including on phones and in the browser.
Model size4.53M parameters100M active parameters
AccessDownloadable ONNX and PyTorch weightsDownloadable publisher checkpoint
Result24 kHz mono speech in Modern Standard ArabicSynthesized speech
Operating limitsIncludes one synthetic Modern Standard Arabic voice. The publisher reports about 200× real-time generation on two CPU threads.Not published
SourcesModel weights · Additional documentationModel weights
InputsTextText, Audio
WeightsOpen weights · Apache-2.0Open weights · Apache-2.0
ReleasedNov 21, 2026Nov 9, 2026