Read Gemini CLI Aloud — Hear Google's AI Coding Agent
Select any text in your Terminal.app where Gemini CLI is running — a plan, a reasoning trace, a code explanation, a diff summary. A floating ▶ appears next to your selection. One click — natural voice with a highlight that tracks each word on the real terminal.
macOS native · Free · No login · 40+ languages · Native terminal AX, no OCR
~/project $ gemini
I scanned the repository and found five usages of the old API client. The recommended migration path is to switch each one to the new retry-aware client, preserving the existing timeout settings. I'll apply the changes in small batches so you can review each one independently.
Highlight follows each spoken word directly inside your Terminal.app window.
One App Covers Every AI Coding Tool on Mac
Same gesture, same word-level highlight — across every IDE and CLI developers actually use.
VS Code
Microsoft's editor
Cursor
AI-native IDE
Xcode
Apple's IDE
Windsurf
Codeium AI IDE
Claude Code
Anthropic CLI
Gemini CLI
Google AI CLI
Codex
OpenAI coding CLI
Copilot CLI
GitHub gh copilot
Aider
AI pair programmer
Three Steps — Clean Terminal AX Path
Gemini CLI runs in Terminal.app. Terminal.app is a native AppKit app, so CastReader reads output through the Accessibility API.
Install
Download CastReader for Mac (.dmg). Grant Accessibility permission. No Screen Recording needed for Terminal.app.
Select
Drag-select any text in Terminal.app while Gemini CLI is running — a plan, a reasoning trace, a tool-call explanation, an error.
Click ▶
A play button appears. One click — audio starts and the highlight glides across the terminal output.
When Gemini CLI Users Add Read-Aloud
Gemini 2.5 produces verbose reasoning — long plans, multi-step rationales, detailed explanations. Listening offloads it so your eyes stay on the diff.
Hear Gemini's plan before applying
Before edits, Gemini CLI outlines its plan. Listening catches 'wrong framework assumption' or 'wrong file targeted' faster than reading.
Listen to reasoning traces
Gemini's extended reasoning can span a dozen paragraphs. Audio lets you absorb the logic while you stretch or sip coffee.
Audio-scan diff summaries
After applying edits, Gemini summarizes the diff. Select, press ▶, verify against the actual diff in another pane.
Review debugging explanations
When a test breaks and Gemini explains why, audio helps you process the root-cause chain without rereading the stack trace.
Study Gemini's coding style
Gemini justifies design choices in prose. Listening on replay embeds the idiom better than reading once.
Accessibility for long agent sessions
A 4-hour Gemini session produces thousands of words of prose. Listen while your eyes focus on the code surface.
Why CastReader Fits Gemini CLI Better Than `say`
macOS's built-in `say` reads a pasted string with no highlight, no word-level sync, and 1990s-era voice quality. CastReader reads your actual selection on the terminal with modern voices.
Word-level highlight on Terminal.app
A transparent overlay tracks each spoken word on the real terminal surface — not a clone pane or side panel.
40+ Kokoro-quality voices
English, Chinese, Japanese, 40+ more — auto-detected per selection. Modern TTS, not 2000s system voices.
Adjustable rate + click-to-jump
1.5× for verbose plans, 0.8× for hard rationales. Click any paragraph to jump — impossible with `say`.
Zero cost, zero signup
No account. No daily word limit. Google's Gemini subscription stays the same — CastReader is the free voice.
Common Questions
How do I read Gemini CLI output aloud on a Mac?▾
Install CastReader for Mac, grant Accessibility permission, run Gemini CLI in Terminal.app, then select any text — a plan, a reasoning trace, an error — and click the floating ▶. Word-level highlight appears on the terminal.
Does it work in iTerm2 / WezTerm / Warp?▾
Terminal.app has the cleanest AX path. iTerm2 works through AX with some quirks. WezTerm uses a custom renderer without AX (unsupported). Warp uses its own text system (partial support).
Will it read markdown in Gemini's output?▾
Gemini CLI renders markdown as ANSI-styled text in the terminal. CastReader reads the underlying characters and skips ANSI escape sequences — the prose comes through clean.
Can I read Chinese Gemini prompts aloud?▾
Yes. CastReader auto-detects CJK and switches to native Mandarin, Japanese, or Korean voices. Mixed-language Gemini replies switch voice at sentence boundaries.
Does it interfere with Gemini CLI's input?▾
No. CastReader only triggers on selection + click ▶. Your terminal input, Ctrl-C, and Gemini CLI's prompt handling stay untouched.
What about running Gemini CLI in VS Code's integrated terminal?▾
VS Code's integrated terminal uses a different AX path — CastReader uses the Electron OCR path there. Works, but needs Screen Recording permission. For cleanest experience, run Gemini CLI in Terminal.app.
Does my prompt leave the machine?▾
Only your selection is sent to the TTS service for synthesis. Nothing is stored beyond the playback session. For fully local, v2 will ship on-device Kokoro.
Is it really free?▾
Yes. 100% free — no account, no credit card, no premium voice gate, no daily word limit. Download the .dmg and use it forever.
Ready to Hear Gemini CLI Speak?
Install once. Run Gemini CLI in Terminal. Select any output. Click ▶. Listen.