Show HN: EdgeWhisper – On-device voice-to-text for macOS (Voxtral 4B via MLX) https://ift.tt/1TfvFxH

Show HN: EdgeWhisper – On-device voice-to-text for macOS (Voxtral 4B via MLX) I built a macOS voice dictation app where zero bytes of audio ever leave your machine. EdgeWhisper runs Voxtral Mini 4B Realtime (Mistral AI, Apache 2.0) locally on Apple Silicon via the MLX framework. Hold a key, speak, release — text appears at your cursor in whatever app has focus. Architecture: - Native Swift (SwiftUI + AppKit). No Electron. - Voxtral 4B inference via MLX on the Neural Engine. ~3GB model, runs in ~2GB RAM on M1+. - Dual text injection: AXUIElement (preserves undo stack) with NSPasteboard+CGEvent fallback. - 6-stage post-processing pipeline: filler removal → dictionary → snippets → punctuation → capitalization → formatting. - Sliding window KV cache for unlimited streaming without latency degradation. - Configurable transcription delay (240ms–2.4s). Sweet spot at 480ms. What it does well: - Works in 20+ terminals/IDEs (VS Code, Xcode, iTerm2, Warp, JetBrains). Most dictation tools break in terminals — we detect them and switch injection strategy. - Removes filler words automatically ("um", "uh", "like"). - 13 languages with auto-detection. - Personal dictionary + snippet expansion with variable support (, ). - Works fully offline after model download. No accounts, no telemetry, no analytics. What it doesn't do (yet): - No file/meeting transcription (coming) - No translation (coming) - No Linux/Windows (macOS only, Apple Silicon required) Pricing: Free tier (5 min/day, no account needed). Pro at $7.99/mo or $79.99/yr. I'd love feedback on: 1. Would local LLM post-processing (e.g., Phi-4-mini via MLX) for grammar/tone be worth the extra ~1GB RAM? 2. For developers using voice→code workflows: what context would you want passed to your editor? 3. Anyone else building on Voxtral Realtime? Curious about your experience with the causal audio encoder. https://edgewhisper.com March 13, 2026 at 11:57PM

Comments

Popular posts from this blog

Complete Guide to E-Commerce Business: Meaning, Models, and How to Start

Micro Niches: The Secret Weapon for SaaS Startups Struggling to Gain Traction

"From Micro Niche to Money Maker: How I Validated My E-Commerce Idea with AI (No Budget Needed)" Published: September 23, 2025 Keywords: Micro niche, AI validation, e-commerce, free tools, startup strategy Introduction Ever wondered if your e-commerce idea is worth pursuing? In this post, I’ll walk you through how I used free AI tools to validate a micro niche, build a lean store, and test demand—without spending a dime. If you’re stuck between ideas or afraid of wasting time and money, this guide is your shortcut to clarity. Step-by-Step Breakdown 1. Finding the Micro Niche Used ChatGPT to brainstorm underserved product categories. Cross-referenced with Google Trends and AnswerThePublic to check search interest. 2. Validating Demand Leveraged Perplexity AI to analyze competitors and market gaps. Ran polls using Typeform and Twitter/X to gauge interest. 3. Building the Store Created a free storefront using Shopify Starter and Canva for branding. Used Durable.co to generate landing page copy in minutes. 4. Driving Traffic Scheduled posts with Buffer across Instagram, Threads, and LinkedIn. Used Notion AI to draft blog content and email sequences. 5. Tracking Results Monitored engagement with Google Analytics and Hotjar. Adjusted product positioning based on feedback from Tally Forms. Key Takeaways Micro niches are goldmines when paired with smart AI validation. You don’t need a budget—just the right tools and strategy. Testing before investing saves time, money, and frustration. Thinking of launching your own store? Drop your niche idea in the comments and I’ll help you validate it with AI—free of charge!