REDDIT · REDDIT// 32d agoOPENSOURCE RELEASE

Echo-TTS MLX brings voice cloning to Apple Silicon

Echo-TTS MLX is an open-source MLX port of Jordan Darefsky’s 2.4B Echo-TTS diffusion model, bringing native Apple Silicon text-to-speech and voice cloning to M-series Macs without PyTorch or CUDA. The repo adds quantization, quality presets, blockwise generation, and both CLI and Python APIs, making a previously GPU-heavy workflow much more accessible to local AI builders on macOS.

// ANALYSIS

This is exactly the kind of port that matters: not a brand-new model, but a practical unlock that puts a serious voice-cloning stack onto consumer Apple hardware. For AI developers, the win is less “new SOTA” and more “real local usability.”

–Native MLX inference means Mac developers can run a 2.4B diffusion TTS system locally instead of treating advanced voice cloning as CUDA-only territory
–The added quantization modes and presets make the project more than a straight port; they turn it into a tunable inference toolkit for latency, memory, and quality tradeoffs
–Blockwise generation and continuation support hint at better streaming and longer-form workflows, which are more useful in apps than one-shot demos
–The included CLI and Python API lower the barrier for experimentation, scripting, and integration into local AI pipelines
–Developers should note the inherited non-commercial licensing constraints around model weights and audio outputs before building products on top of it

// TAGS

echo-tts-mlxspeechaudio-genopen-sourceinference

DISCOVERED

32d ago

2026-03-10

PUBLISHED

36d ago

2026-03-07

RELEVANCE

8/ 10

AUTHOR

mattznoj