I have real respect for Talon — it lets people run an entire computer, and write actual code, without touching a keyboard, and for many it's the difference between working and not. I build Talkie, which is a far smaller idea, and being honest about that gap is the fastest way to help you choose. Talon is a commitment you learn; Talkie is meant to disappear.
Short version: Talon controls your entire computer by voice. Talkie doesn't. Talkie is on-device dictation and capture for Mac and iPhone that turns your speech into structured, searchable data — and then pipes it into agents like Claude Code. If you want a full voice-control OS layer, use Talon. If you want to talk instead of type and feed that text to your tools, use Talkie. They overlap a little around "voice + code," but they're doing different jobs.
What Talon does well
Talon is a hands-free computer-control system, and it's genuinely powerful.
- Full computer control. Voice commands, mouse movement, clicking by noise (pops and hisses), and optional eye tracking. You can operate the entire machine without a keyboard or mouse.
- Voice coding grammars. Talon's command sets let you speak code — symbols, snippets, cursor motions, refactors — using spoken grammars. The community command set (talonhub/community) is large and mature.
- Accessibility. For people who can't use a keyboard or mouse due to RSI, injury, or disability, Talon is often the difference between working and not. This is where it's irreplaceable.
- Deep customization. Talon embeds CPython. You script your own commands, contexts, and grammars in Python and reshape almost anything.
- On-device. It ships a free speech engine (with Conformer in beta) and works with Dragon; recognition runs locally on macOS, Windows, and Linux. There's a free public version plus a Patreon patron/beta tier for early features.
The trade-off is real: Talon has a steep learning curve. You're learning a command language, wiring up grammars, and maintaining config. That's the price of controlling everything by voice.
Where Talkie is different
Talkie isn't trying to control your computer. It's trying to make dictation dead simple and then get the text out of the way and into your tools.
- Simplicity, dictation-first. Talk, get text. There's no command grammar to memorize and no config to maintain. You're dictating, not programming a voice OS.
- Local-first and programmable. Captures live locally, and the CLI makes them available to scripts, workflows, and agentic tools.
- Local recognition and durable capture. Talkie offers local Parakeet and Whisper paths on Mac and Apple Speech or Parakeet paths on iPhone. Captures enter a structured local library rather than a voice-command grammar.
- Structured captures, not just a text stream. Every dictation becomes a searchable, structured capture — local knowledge you can query, not a transient paste.
- A CLI for your scripts and agents. Talkie ships
@talkie/cli, which exposes your captures as structured data for pipelines:
bun add -g @talkie/cli
- Dictate into Claude Code. Because captures are structured and scriptable, Talkie acts as a remote control for your agents — you talk, and the text drives an agentic CLI like Claude Code instead of you typing prompts.
- Give agents visual context. Add a full screenshot, selected screen region, or camera capture to a spoken instruction, so your agent gets the thing you see as well as the words you say.
- Cross-device. macOS, iPhone, and Apple Watch, synced over iCloud. Capture a thought on your watch, use it on your Mac.
The honest limitation: Talkie is Apple-only, and it won't move your cursor or let you code hands-free. It's dictation and capture, not computer control.
Side by side
Dictation without a command system
- Quick hold-to-talk
- Durable capture history
- Voice input for agents
Control the computer hands-free
- Cursor, clicks, and commands
- Voice coding grammars
- Python customization
| Decision factor | Talkie | Talon Voice |
|---|---|---|
| Best at | Dictation + capture | Hands-free computer control |
| Dictation | The primary interaction | One part of a control system |
| Full control | Not an accessibility layer | Cursor, clicks, keyboard, apps |
| Customization | CLI, SDK, workflows | Python + command grammars |
| Output | Speech captured as text/data | Deterministic commands |
| Desktop | macOS | Mac, Windows, Linux |
Which one should you pick?
Pick Talon if you need to operate your computer without your hands — for accessibility, for RSI recovery, or because you want to write code entirely by voice. Nothing here matches it for full hands-free control, and if that's your requirement, the learning curve is worth it. Talon is the right tool for that job.
Pick Talkie if what you actually want is to stop typing so much: on-device dictation whose output isn't just pasted text but structured captures you can search, script, and pipe into your agents. If your day is "talk out a prompt, let Claude Code do the work," Talkie is built for exactly that, and it works the same way across your Mac, iPhone, and Watch.
They're not really competitors. Talon is a voice-control OS layer. Talkie is dictation that feeds your agents. Plenty of people would be well served by either — or both.
Download Talkie for Mac and iPhone →
Related reading: