Voice Commands: Hands-Free AI Interaction for Developers

Keyboards and mice have been the primary interface for software development since the beginning. But there are moments when typing is not the most efficient input — when you are looking at code and want to ask a question without switching context, when your hands are busy with a physical setup, or when you want to dictate a longer explanation faster than you can type it. Voice commands add a parallel input channel that complements your existing workflow rather than replacing it.

Gonu AI's voice command system is built into the desktop application. It uses speech-to-text to capture your voice and converts it into actions the AI agent can execute. This is not a limited set of predefined commands — you can speak naturally, and the AI interprets your intent.

Natural Language Voice Input

Unlike traditional voice assistants that recognize a fixed vocabulary of commands, Gonu AI's voice system accepts natural language. You can say "add error handling to the database connection function" or "what does this useEffect hook do" or "run the tests for the authentication module" — the AI understands the intent and acts accordingly.

The speech-to-text engine runs with low latency, transcribing your voice in real time as you speak. The transcription appears in the chat input, and when you finish speaking, the AI processes it just as if you had typed the message. This means you get the full power of the coding agent — file access, code generation, terminal execution, and workspace search — all through voice.

Use Cases for Voice in Development

The most common use case is asking questions while reading code. When you are deep in a file trying to understand a complex function, switching to the chat to type a question breaks your focus. With voice, you say "explain what the reconciliation loop in line 47 does" while keeping your eyes on the code. The AI reads the file, finds line 47, and explains the logic.

Another frequent use case is dictating commit messages and documentation. Developers often write terse, unhelpful commit messages because typing a detailed description feels like a chore. Speaking the description — "this commit adds retry logic to the payment webhook handler to handle transient Stripe failures" — takes five seconds and produces better documentation.

Voice is also useful during code reviews. As you scroll through a pull request, you can dictate comments: "the variable name here is misleading, it should be called remaining count instead of total" — and the AI captures it as a review comment you can post.

Voice During Meetings

The voice system is also active during meeting mode. You can ask the AI questions while on a call — "what was the action item from the last standup" or "summarize the last five minutes" — and get answers in the overlay without the other participants hearing anything. The AI responds visually in the stealth overlay while your microphone stays muted or directed at the meeting.

Speech-to-Text Providers

Gonu AI supports multiple STT providers so you can choose the one that best fits your latency, accuracy, and privacy needs. The supported providers include Deepgram for low-latency real-time transcription, OpenAI Whisper for high-accuracy transcription, and local models via Ollama for fully offline voice processing. You configure the provider in the settings, and the voice system uses it for all transcription.

Privacy Considerations

Voice data is handled the same way as all other data in Gonu AI — it goes only to your configured STT provider and is not stored on Gonu AI's servers. If privacy is a concern, you can use a local STT model that processes voice entirely on your machine without any network requests.

Getting Started

Voice commands are available on all plans. Download Gonu AI, configure your preferred STT provider in settings, and use the microphone button or keyboard shortcut to start speaking. The AI listens, transcribes, and executes — you just talk and code.