DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to speed up Kokoro on an Apple Silicon Mac with Kokoro

Enable MPS fallback so Kokoro uses the Mac GPU.

Short answer: Install Kokoro with pip; then write your Kokoro script, for example run-your-kokoro-script.py; run it with PYTORCH_ENABLE_MPS_FALLBACK=1 python run-your-kokoro-script.py. All steps.

  • Content
  • Beginner
  • 15 minutes
  • From the README
  • Claude Code
  • Codex
  • Cursor
  • Gemini CLI

platform. This tip applies only to Apple Silicon Macs.

Who it is forA Mac user generating long audio locally.

The problemCPU-only generation is slow for long scripts.

The outcomeFaster local generation on M1 to M4 Macs.

What you need first

  • Python with pip (the README conda file uses Python 3.9)
  • espeak-ng installed for fallback and some languages
  • Optional: an Apple Silicon Mac for MPS acceleration

Steps

  1. Install Kokoro with pip
  2. Write your Kokoro script, for example run-your-kokoro-script.py
  3. Run it with PYTORCH_ENABLE_MPS_FALLBACK=1 python run-your-kokoro-script.py
  4. If dependencies conflict, use the conda environment.yml from the README

Prompt for your agent

Using hexgrad/kokoro on my Mac, set up a script and run it with `PYTORCH_ENABLE_MPS_FALLBACK=1 python run-your-kokoro-script.py`, then compare timing with and without the variable.

Caveat. This card is written from the hexgrad/kokoro README, not from a test run by Deduped. Before you start, make sure you have Python with pip. Check the repo for newer instructions before you rely on these steps.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for hexgrad/kokoro and "Speed up Kokoro on an Apple Silicon Mac"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer