Skip to main content
← Back to sources

KittenTTS Releases Three New Ultra-Compact Text-to-Speech Models Under 25MB

Published 2026-03-25Foundation ModelsLow⭐ Timeline Candidate

Summary

KittenML has released three new text-to-speech (TTS) models under the KittenTTS project on GitHub, with the smallest model weighing in at less than 25MB. The project appears to be an open-source effort focused on creating extremely lightweight TTS models suitable for edge deployment and resource-constrained environments. The emphasis on minimal model size (sub-25MB) positions these models for on-device and embedded use cases where bandwidth, storage, and compute are limited. While the article c

Alignment: Neutral
Related Positions: ai-infrastructure-strategy.md
text-to-speechttssmall-modelsedge-aiopen-sourcemodel-compressionkitten-ttslightweight-inference
KittenTTS Releases Three New Ultra-Compact Text-to-Speech Models Under 25MB — Intelligence — Agentic Developer Tools Radar · Signal