Long replies without typing
For WeChat, Feishu, Slack, or any chat app: speak first, then make small edits before sending.
Ousia Voice for Mac is available now · On-device-first voice typing and file transcription
Download
デスクトップ向けのローカル音声入力とファイル文字起こし。
グローバルショートカットで話し、認識結果を現在のアプリへコピーまたは貼り付けできます。音声・動画ファイルの文字起こしにも対応します。
12ms fast response*
No account
On-device first
Audio and video transcription
Current app
Mail
You say
Can you help me reply to this email and say I can take a look this afternoon, and if it works for them we can do a quick sync tomorrow morning.
Press the shortcut and speak
Transcribing on device
Ousia Voice enters
I can take a look this afternoon. If that works for you, we can do a quick sync tomorrow morning.
Speak where you are already writing.
Start documents by speaking
In Mail, Notion, Pages, or Notes, turn a thought into editable text first.
Tell AI more context
In Cursor, VS Code, or the browser, dictate prompts, comments, and commit messages.
Capture it when it arrives
When a thought appears, use your voice before your hands catch up.
Long-form input does not have to start at the keyboard.
Ousia Voice is not another editor. It helps you get the first draft of text inside your existing workflow.
Think · Type · Delete · Type again
Shortcut · Speak · Return to current app · Tweak
For people who work with words all day.
Speak a full thought first, then refine it into an article, script, or email.
Turn class recordings, reading notes, and assignment ideas into text drafts.
Dictate requirements, prompts, comments, and commit messages directly into your editor.
Turn videos, podcasts, and interview material into searchable text for editing.
Convert recordings into text so they are easier to search, copy, and organize.
Start on device by default, without sending your voice to an online account first.
Short input and long files in one desktop tool.
Press a shortcut, speak, and enter text
For messages, emails, documents, prompts, and quick notes. Text can return to the current app, reducing copy and paste.
Mouse key 5
Listening
Turn audio and video into text first
For meetings, interviews, courses, and video material. Generate text or draft subtitles before review and editing.
meeting_audio.m4a
Draft subtitles
Not every speech-to-text tool belongs on your desktop all day.
Dimension
Common options
Ousia Voice
Response
Usually requires opening dictation, a web page, or a project first.
Shortcut trigger, recording state in 12ms.*
Where text goes
Often transcribes in a separate window, then you copy it back.
Text returns to the current app: chat, docs, notes.
Accuracy tuning
Repeated recognition mistakes often need manual fixes.
Maintain common replacements for names, terms, and fixed phrases.
File transcription
Short input and file transcription are often separate tools.
Dictation and audio/video transcription live in one desktop app.
Privacy control
Accounts, uploads, and online processing are often the default path.
Starts on-device by default; online features are optional.
標準でローカル処理。
標準で音声をアップロードしません
音声認識はローカルで実行され、声をクラウドへ送りません。
標準で文章を書き換えません
Ousia Voice は文章作成支援ではなく、元の言葉を保つことを目指します。
標準でアカウント不要
入力レイヤーを先に磨き、商用サービスは最小限に保ちます。
A real Mac desktop app, not another web text box.
Pay once. No subscription.
Ousia Voice Pro uses a clear one-time price. The current version is for macOS; future desktop versions will follow the release plan.
まずは Mac で試す。
Ousia Voice は Mac App Store から入手するか、macOS DMG インストーラを直接ダウンロードできます。
直接ダウンロード
macOS DMG
Mac App Store version in preparation.
Ousia Voice について、まず知っておきたいこと。
Ousia Voice では何ができますか？
話した内容を文字にして現在のアプリへ戻せます。録音や動画をテキストや字幕の下書きにすることもできます。
Ousia Voice にアカウントは必要ですか？
不要です。開いてすぐ使えます。録音と文字起こし履歴は標準では保存されません。
Ousia Voice は標準でオンライン処理しますか？
標準ではまずデバイス上で処理します。オンライン機能が必要な場合は、設定で有効にできます。
今どこでダウンロードできますか？
このページから macOS DMG を直接ダウンロードできます。Mac App Store 版が公開された後も、別のダウンロード入口として残します。
* 12ms refers to the interface response time from shortcut trigger to recording-state activation in a test environment. It does not represent full speech transcription time. Tested on an Apple Silicon Mac while idle, using the better result from repeated triggers; actual performance varies by device, system load, audio length, and recognition settings.