Claude Code just shipped /voice — voice input directly in the terminal.
I tested it the moment it landed. Said "Claude Code." It transcribed "clot coat."
Not great. But let's be fair about what /voice actually is today.
What /voice does
- Voice input only — you speak, it transcribes to text, Claude responds in text
- Terminal CLI only — no VSCode support yet
- No voice output — Claude doesn't speak back
- No vocabulary learning — it will keep getting "Claude Code" wrong
That last point is the real issue for daily use.
The comparison
| Feature | /voice | ElevenLabs | Wispr Flow |
|---|---|---|---|
| Voice input | ✅ | ❌ | ✅ |
| Voice output (TTS) | ❌ | ✅ | ❌ |
| Works in terminal | ✅ | ✅ | ✅ |
| Works in VSCode | ❌ | N/A | ✅ |
| Works everywhere on Mac | ❌ | N/A | ✅ |
| Vocabulary learning | ❌ | N/A | ✅ |
Why vocabulary learning matters
Wispr Flow remembers every correction. Fix "Claude Code" once, it's correct forever. Same for your project names, framework abbreviations, and technical jargon.
It works in every input field on Mac — Slack, browser, terminal, editors, everything. And it gets smarter on YOUR vocabulary with every use.
The verdict
/voice is a promising v1. If you live entirely in the terminal and don't mind re-correcting the same words, it works.
But for daily developer workflows on Mac, Wispr Flow is still in a different category. The vocabulary learning alone makes it irreplaceable.
Has anyone found a workflow where /voice actually outperforms dedicated tools? I'm genuinely curious.
Top comments (0)