Hold a key, speak naturally, release — your words stream directly into any text field. Cloud AI-powered voice input for macOS with context awareness and zero local dependencies.
VIEW ON GITHUB ↗SpeakMore is a lightweight macOS menu bar app that converts speech to text using cloud-based multimodal AI models. Unlike traditional speech-to-text tools that chain separate recognition and enhancement stages, SpeakMore sends your audio directly to a multimodal AI model in a single step — the model handles recognition, punctuation, formatting, and domain-specific terminology simultaneously. Text streams in real-time into whatever text field you're focused on, whether it's a code editor, email client, chat app, or document. With context awareness that captures your current app, window title, and document path, plus short-term and long-term memory of your speech patterns, SpeakMore adapts to your vocabulary and writing style over time.
Audio goes directly to a multimodal AI model — no separate speech recognition and text enhancement stages. The model handles transcription, punctuation, formatting, and domain-specific terminology in a single pass for faster, more accurate results.
Text streams directly into the focused text field as it's generated — via macOS Accessibility API, keyboard simulation, or clipboard injection. Three insertion methods with automatic fallback ensure text arrives in any application.
SpeakMore captures your current app name, window title, and document path before each transcription. This real-time context helps the AI model produce more accurate, contextually appropriate text.
After 10 utterances, SpeakMore builds a short-term context snapshot with your recent topics and vocabulary. Daily profiles capture your identity, primary domains, language habits, and frequently used proper nouns for increasingly personalized transcription.
Set global or per-app transcription instructions — from faithful transcription to structured output. Add a priority terminology list for proper nouns, technical terms, and domain-specific vocabulary that the AI should always get right.
Choose from Google Gemini, DashScope (Qwen), OpenRouter, or any OpenAI-compatible endpoint. Bring your own API key — no subscription, no vendor lock-in. Switch models with a single click.
No model downloads, no local GPU required. The entire app is ~2MB. Pure cloud-based architecture means you always get the latest AI models without updating the app.
Press and hold the hotkey (default: Fn key). SpeakMore captures your current app context and starts recording from your microphone.
Speak naturally — the audio is recorded at 16kHz and a real-time equalizer shows your voice level. Say what you mean, in any language.
Release the key. Your audio is sent to the cloud AI model and transcribed text streams directly into the focused text field in real-time.
Write commit messages, code comments, documentation, and Slack messages by voice. Context awareness ensures technical terms and variable names are transcribed correctly.
Draft articles, emails, and social media posts at the speed of speech. Long-term memory adapts to your writing style and frequently used phrases.
Dictate meeting notes, reports, and project updates directly into any application. Per-app prompts let you tailor output format for different tools.
Speak in any language the AI model supports. SpeakMore handles mixed-language input naturally — switch between English and Chinese mid-sentence.
| Feature | SpeakMore | Traditional Voice Input |
|---|---|---|
| Architecture | One-step multimodal AI | Separate STT + enhancement |
| Context awareness | App, window, document + memory | None |
| Custom terminology | Per-app prompts + glossary | Limited or none |
| Model updates | Always latest cloud models | Manual updates required |
| Local storage | ~2MB, no models | Gigabytes of models |
| Open source | MIT license | Proprietary |
SpeakMore is free and open source. Clone from GitHub, build with Xcode, and start dictating.
VIEW ON GITHUB ↗