AI/ML●●●Banger
Rapid-MLX – Run local LLMs on Mac, 2-3x faster than alternatives
Claims 4.2x Ollama speed with 0.08s cached TTFT on Apple Silicon.
WizardrySolve My Problem
raullen
941mo ago
Back to browse
GitHub Repository
OpenClaw local TTS plugin powered by mlx-audio, zero API key, zero cloud dependency
14 starsTypeScript
Local TTS for OpenClaw on Apple Silicon (MLX-Powered, Zero Setup)
by ZacharyZZ·Feb 20, 2026·2 points·0 comments
AI Analysis
●●SolidNiche GemCozy
MLX-powered local TTS plugin for OpenClaw—elegant but audience is Apple Silicon only.
Strengths
- •Zero setup: bootstrap uv, download model, auto-restart on crash
- •OpenAI endpoint compatibility means drop-in replacement
- •Multi-language support (Kokoro, Qwen3) with 3-sec voice cloning option
Weaknesses
- •Locked to Apple Silicon M1+; no Windows/Linux/Intel support
- •Audience is intersection of OpenClaw users + macOS—very narrow
- •Existing alternatives (Edge TTS in OpenClaw, Self-hosted GPU servers) already work
Category
Target Audience
Apple Silicon Mac users running OpenClaw; developers needing private/offline TTS
Similar To
OpenClaw built-in Edge TTS · oobabooga local TTS · Bark
Post Description
I built an OpenClaw plugin that runs text-to-speech entirely on your Mac. No API keys, no cloud, no pre-installed Python required.
It wraps mlx-audio and handles the full lifecycle: bootstraps its own Python environment via uv, downloads the model on first run, manages the server process, auto-restarts on crash, and exposes a standard OpenAI-compatible /v1/audio/speech endpoint.
Installation:
openclaw plugin install @cosformula/openclaw-mlx-audio Four models out of the box:
• Kokoro-82M: ~400 MB RAM, fastest, good for English/Japanese • Qwen3-TTS-0.6B: ~1.4 GB RAM, best Chinese quality, 3-second voice cloning • Qwen3-TTS-1.7B VoiceDesign: generate voices from text descriptions • Chatterbox: 16 languages, ~3.5 GB RAM
Works on 8 GB Macs with Kokoro or Qwen3-0.6B. A proxy layer injects model-specific parameters so OpenClaw's TTS client needs zero changes.
Why not just run mlx-audio directly? You can. This plugin removes the setup friction: no Python version juggling, no pip install, no manual server management. It also adds OOM detection, memory pre-checks, startup progress tracking, and hot config reload.
Similar Projects
AI/ML●●Solid
MLX-Ruby – Ruby Bindings for Apple's MLX ML Framework
Full MLX power in Ruby: lazy arrays, Metal GPU, transformer layers—but Ruby adoption risk.
WizardryDark HorseNiche Gem
skryl
113mo ago
AI/ML●●●Banger
I built a free CharacterAI that runs locally
Free local CharacterAI with voice cloning under 10s audio, plus ESP32 hardware integration.
Zero to OneWizardrySolve My Problem
akadeb
842mo ago
AI/ML●●●Banger
Apfel – The free AI on your Mac
Unlocks Apple's locked LLM with OpenAI-compatible server for existing SDKs.
Zero to OneSolve My ProblemBig Brain
franze
7431572mo ago
AI/ML●●●Banger
SwiftLM – Qwen Chat on iPhone, 100B+ Moe on M5 Pro 64GB (Native Swift)
Native Swift inference with SSD streaming runs 100B MoE models without kernel panics.
WizardryNiche Gem
aegis_camera
122mo ago
Data●●Solid
Benchmarking Apple Silicon unified mem for GPU-accelerated SQL analysis
The repo does one practical thing well: quantify the real-world impact of Apple Silicon's unified memory on analytics by running six TPC-H queries plus a GPU-favorable QX and shipping the raw charts and code. It's specific and empirical — you get MLX vs NumPy vs DuckDB numbers and PNGs, not just hand-wavy claims — but it's narrowly scoped to M4 hardware and small-ish scales, so its conclusions are useful for experimentation rather than sweeping generalization.
WizardryNiche Gem
sadopc
313mo ago