Voice Input#

Most AI tools accept text input, but for many tasks — especially explaining processes, thinking out loud, or capturing quick notes — speaking is faster and more natural than typing. Voice input reduces the friction of getting your thoughts into the AI, and tends to produce more conversational, less self-edited raw material.

When to Use#

• You’re documenting a process you know well and can explain faster than you can type. See the SOP / Process Documentation playbook.
• You want to capture a quick journal entry or reflection without switching to a keyboard. See the Automated Journaling playbook.
• You’re working through a problem out loud and want the AI to engage with your thinking in real time.
• You’re on the go and want to capture something before you lose it.

Platform Voice Support#

Claude#

• Mobile app (iOS/Android) — Full voice mode available on all plans (free and paid). Supports voice input and output with multiple voice personalities. English only. Tap the microphone icon to start.
• Claude Code CLI — Voice dictation via the /voice command. Push-to-talk: hold spacebar, speak, release. Requires v2.1.69+ and a Claude.ai account. Currently in limited rollout. Supports 20 languages.
• Desktop / Web — Voice mode is in gradual rollout and not yet broadly available.

ChatGPT#

• Mobile app (iOS/Android) — Advanced Voice Mode uses GPT-4o to process audio directly. Supports video and screen sharing during voice conversations. Available to Plus, Pro, and Team users; free users get a monthly preview.
• Web app (chatgpt.com) — Voice mode available to all logged-in users.
• Windows desktop app — Voice mode supported.
• macOS desktop app — Voice mode was retired in January 2026.

Ollama (Local Models)#

Ollama has no native voice support. Voice input requires combining a speech-to-text tool with Ollama:

• Typical stack: Whisper (STT) → Ollama (LLM) → pyttsx3 or Bark (TTS)
• Notable open-source projects:
  • ollama-voice — Whisper + Ollama + pyttsx3, fully offline
  • local-talking-llm — Whisper + Ollama + Bark, no internet required
• All options require manual setup (Python, model downloads, etc.)

Tips#

• Talk like you’re explaining something to a colleague, not dictating a document. The AI can clean up the structure later.
• Don’t worry about filler words, restarts, or imperfect grammar — transcription tools and AI are both good at handling natural speech.
• For longer explanations, pause between logical sections. This makes it easier to review and edit the transcript afterward.