Knows your words
Names, product terms, jargon. Ore picks them up as you use it, or you can add them to your personal dictionary yourself.
Ore 6.5 · Voice dictation for Mac and Windows
Say it messy.
Get it structured.
Ore turns rambling speech into structured text. Fillers gone, corrections resolved, points split into lists, right at your cursor. Tuned for coding agents like Claude Code and Cursor.
Free to start · No API key needed
Vibe Coding · Agent Coding
Tell your agent once
Tuned for coding agents like Claude Code, Codex and Cursor. Think out loud. Ore drops the filler, keeps only what you settled on, spells the tech terms right and hands your agent a clear, itemized instruction.
uh so split up the settings page, like the react component is way too long, make it three, and use zustand for state, no wait, skip zustand for now, just use context, and and add vitest unit tests, then open a p r against dev
Sent to Claude CodeSplit up the settings page:
- SettingsPage is too long; break it into three child components
- Use React Context for state for now; don’t add Zustand yet
- Add Vitest unit tests
- Open a PR against the dev branch when done
Filler removedKept your final callTerms spelled rightSplit into steps
Talk straight into the coding agents you use
- Claude Code
Codex
Cursor
GitHub Copilot
Windsurf
Gemini CLI
Trae
Cline
300+ AI and developer terms built in: Claude Code, MCP, Subagent, Supabase, Next.js, Prisma… File names, commands and error codes come out as written.
Every app
If you can type there, you can talk there
Tap Fn, speak, tap again. The text goes straight to your cursor. No plugins, no switching windows. Pick any shortcut you like.
- Slack
- Gmail
- Notion
- ChatGPT
- Claude
- VS Code
- Cursor
- Linear
- Google Docs
- iMessage
- Outlook
- Figma
The details are handled
Switch languages mid-sentence
English, Mandarin, Cantonese, Japanese, and Korean, detected automatically. Mixing English and Chinese in one sentence works too.
Talk it into a note
Hit the note shortcut and your voice becomes a note. Clean it up, give it structure, then export TXT, Markdown, or PDF.
Always know where it is
Teal means listening, amber means cleaning up, red means that pass failed. Esc exits any time.
Beyond dictation
Done talking? Keep going.
While recording, save it as a note, hand it to the agent, or ask search engines and AI at once. Or open Ore and tell it what today needs.
Send it while you speak
Press an arrow key while recording: ← save as a note, ↑ hand it to the agent, → open parallel search.
Ask from the home screen
The agent takes notes and searches the web.
Download Ore
macOS and Windows are both ready. Sign in and start talking. No API key.
Apple silicon and Intel · macOS 12+ · Windows 10+
FAQ
Which apps does Ore work in?
Anywhere you can type: Slack, Gmail, Notion, Google Docs, your browser, your code editor, and AI tools like ChatGPT and Claude. The text lands at your cursor. No plugins or integrations.
How is Ore different from built-in dictation?
Built-in dictation writes down every word, including every “um”, repeat and change of mind. Ore removes filler words, keeps only your final version, adds punctuation and splits points into lists, and uses your personal dictionary for names and terms, so the text is structured and ready to send.
Is Ore good for vibe coding? Does it work with Claude Code and Cursor?
Yes, it is the scenario Ore is tuned for. Speak your request straight into Claude Code, Codex, Cursor, GitHub Copilot, Windsurf or any coding agent and Ore turns it into an itemized instruction. 300+ AI and developer terms (MCP, Subagent, Supabase, Next.js, Prisma and more) are built in, and file names, commands and error codes come out as written.
Can I try it for free?
Yes. The free plan includes 30 minutes of voice and 100 agent credits a month, enough to try dictation, notes, and the agent before you decide to subscribe.
Do I need an API key?
No. Ore handles speech recognition and cleanup. Download, sign in, and talk.
What happens to the audio?
To recognize and clean up what you said, the audio and the text it needs go to Ore’s cloud. The rules are in the privacy policy.
Which languages does it support?
Mandarin, English, Cantonese, Japanese, and Korean. Auto-detect works, including mixed Chinese and English.
Does it need a connection?
Yes. The models run in the cloud, so you always get the current version.
Which systems does it run on?
macOS 12 or later, Intel and Apple silicon. Windows 10 or later.
What can the agent do?
It writes and edits notes, and searches the web.