Voilà: A macOS Voice Input Tool Built for Coders

Claude-code">A Developer Built His Own Voice Input Tool for Claude Code

An indie developer known as qfdk has released Voilà, a macOS-native voice input tool purpose-built for software development workflows. The tool was born out of frustration with typing lengthy prompts into Claude Code — Anthropic's CLI coding assistant — and the shortcomings of existing voice input solutions.

After 6 months of development, Voilà now handles hundreds of daily invocations in the creator's own workflow. The developer is offering 20 free license keys to celebrate the launch.

Why Existing Tools Fell Short

qfdk, a developer based overseas who regularly mixes Chinese, English, and French in a single sentence, found that most commercial voice input tools couldn't keep up. Common pain points included high pricing, single-language limitations, no text polishing, and — critically — lost characters when pasting into terminal emulators.

'I tried everything on the market,' qfdk explained. 'They either cost too much, only support one language, or break when you paste into a terminal.'

Key Features That Set Voilà Apart

Voilà focuses on practical, developer-centric functionality rather than flashy UI. Its standout features include:

• Multi-engine speech recognition: Supports Soniox, ElevenLabs, Volcano Engine, Groq Whisper, and Apple Speech — users can switch engines based on their needs
• Real-time subtitle overlay: A floating window displays recognized text as you speak, so you can catch errors before committing
• AI-powered text polishing: Automatically removes filler words, adds punctuation, and fixes speech errors — but intelligently skips polishing when confidence is high, saving roughly 200ms of latency
• Terminal-friendly pasting: Uses macOS Accessibility API to paste via menu commands in Ghostty, iTerm2, and Kitty, avoiding the character-dropping issues caused by standard Cmd+V events
• Custom vocabulary: Force-replace specific terms like proper nouns and technical jargon to prevent misrecognition

Solving the Multilingual Developer Problem

The multilingual angle is particularly noteworthy. Developers working across borders — or simply using libraries and APIs with names from different languages — often struggle with voice tools that expect monolingual input. Voilà's multi-engine approach lets users pick the best recognizer for their language mix.

The custom vocabulary feature addresses another common frustration: speech engines mangling technical terms. Without it, a product name like 'Soniox' might get transcribed as gibberish. With the vocabulary table, these terms are force-corrected every time.

Who Is This For?

Voilà targets a specific niche: macOS developers who rely heavily on CLI tools and want to dictate rather than type. The tool is especially useful for those who:

• Spend significant time writing prompts for AI coding assistants like Claude Code or GitHub Copilot CLI
• Work in multilingual environments
• Use terminal emulators as their primary interface

The 20 free licenses are available on a first-come, first-served basis. Interested users can find details through the developer's original post.

The Bigger Picture

Voilà reflects a growing trend of developers building hyper-specific tools around AI-assisted coding workflows. As tools like Claude Code, Cursor, and Aider become daily drivers, the friction of communicating with these tools — typing natural language prompts hundreds of times per day — becomes a real productivity bottleneck. Voice input, done right, could be the missing layer in the AI coding stack.