VoicedJoin the Waitlist

Voiced

Just
speak.

A macOS dictation app built for Traditional Chinese users. Chinese-English code-switching, Taiwanese wording, and full-width punctuation all handled.

Speak, and it's text.

Watch the words appear as you speak, so you never have to guess what the system heard. When you finish, a cleaned-up version lands right at your cursor.

00:00
正在聽寫中EN
我剛剛跟 Claude Code 試了一下 vertex，它真的順很多。
右⌥插入游標位置

Built for Traditional Chinese

Most dictation tools are built for English; everything else is just "supported," never truly tuned. Traditional Chinese usually gets lumped under "Simplified Chinese," so the output mixes in Simplified characters and mainland wording, or mechanically converts to the wrong Traditional form. Voiced flips this around — Traditional Chinese comes first, tuned at every layer from recognition to rewriting.

01

Genuinely native Traditional Chinese

「干擾」 won't become 「乾擾」, and 「皇后」 won't become 「皇後」. An AI model resolves one-to-many Simplified-to-Traditional ambiguity at the source.

02

Taiwanese wording, left alone

軟件 → 軟體, 視頻 → 影片, 教程 → 教學. Even if the raw output drifts into mainland terms, what you paste still reads like Taiwan.

03

Full-width punctuation, completed automatically

Uses full-width punctuation and the Taiwanese punctuation system — no more stray half-width commas or the wrong quotation marks (“”).

Code-switching

Mix Chinese and English,
without it getting weird.

Many voice tools only support one main language and force English brand names, technical terms, and proper nouns into awkward Chinese. Voiced recognizes how people actually talk: English stays English, Chinese comes out Traditional — whatever language you speak, that's what it outputs.

8.4s

我剛剛用 Claude Code 試了一下新的 MCP server，它處理 tool use 比上週順很多，尤其是中英混說， 幾乎不再把 token 拼錯。

Your voice never
passes through our servers.

Voiced uses a direct-audio, ZDR text-routing architecture: audio goes from your Mac straight to the ASR provider; once it becomes text, Voiced servers briefly route it to the polish model. That server-side handling is ZDR, not stored, and never used for training.

See the full data-flow diagram
AudioTextControl / subscription
CLIENT

Your Mac

Voiced.app · local

Audio, dictionaries, and history live locally by default
VENDOR · ASR

Speech-to-text

Zero data retention (ZDR) · no training

Released as soon as it becomes text
VOICED · SERVER

Text polish gateway

ZDR · no content storage · no training

Brief text routing and subscription checks
VENDOR · LLM

Polish

Zero data retention (ZDR) · no training

Returned after rewriting, nothing kept
See the full privacy policy →
Three guarantees

Zero training, zero data retention,
clear accountability.

Zero training

Your voice and text are never used to train any model.

Zero data retention

Providers release processed audio and text; Voiced servers release briefly routed polish text after completion (ZDR).

Clear accountability

Voiced is operated by ASWAY LTD, registered in England and Wales, and built by a Taiwan-based team. We handle personal data under applicable GDPR / UK GDPR and Taiwan Personal Data Protection Act requirements.

Your words, your way

Rules and styles,
switch them and see the difference.

In the app, set your background terms, choose cleanup rules like filler removal, repeat removal, self-correction fixes, and punctuation; then enable rewrite options such as paragraphs, bullets, more specific wording, Markdown, or custom style.

Raw ASR output

嗯欸我跟你介绍一下 voiced 啊它就是 macos 上的 dictation app um 支持中英混说有一个 polish 功能它会改错字移除赘字修正口误举个例子 vector 啊不对是 vertex 还会加标点符号也可以整理成列点我觉得超方便超方便

→
Polished
Let me introduce Voiced: a macOS dictation app built for Traditional Chinese and Chinese-English code-switching. AI Polish cleans up the transcript after speech. This pass removes fillers like um, removes repeated phrases, turns vector, no, vertex into vertex, adds punctuation.
Clean up
Rewrite
Power Mode

When you need more precision,
turn on Power Mode.

Power Mode is off by default. When enabled, Voiced knows which app you are using and can use Voiced transcripts and rewrites created in that app to summarize each app's writing style, so output fits the place you are writing.

Power Mode

Smart word choice

Learns from your corrections and recurring mishearings, so the next transcript is sharper.

App-aware tuning

When enabled, Voiced knows which app you are using and uses that app context to tune tone, terminology, and punctuation rhythm.

Per-app polish stylesBeta

Set a polish preference for each app; Voiced can also summarize transcripts and rewrites created in that app to learn a better output style.

CodexSet it to preserve spoken technical notes, with terms like API, Worker, and Tab left intact.
ChatGPTSet it to a conversational tone, with more natural everyday phrasing and a little personal texture.
ClaudeSet it to structured reasoning, breaking long thoughts into clearer layers before the next step.

Your special terms,
it remembers them all.

Names, brands, acronyms, and technical terms often get misheard? Voiced lets you tune recognition three ways, so it understands what you really meant.

  • Custom vocabularyAdd names, custom nicknames, or special terms — recognized with priority during transcription
    Priority
  • Common mishearingsA word keeps getting misheard? Map "what you said" to "what it's often heard as" for sharper recognition
    Correct
  • Custom replacementsSay a word, expand it into the full content
    Replace
Voiced — Custom dictionary
Custom vocabulary · priority recognition
Claude Code
MCP server
tRPC
Common mishearings
mainoften heard as man
textoften heard as tags
Custom replacements
"my email"→ aaa@gmail.com

One tap is enough.

Every shortcut is customizable. Dictation uses a low-conflict right-side modifier; rewriting selected text uses left Command + Option without taking over any letter keys.

esc
F1
F2
F3
F4
F5
F6
F7
F8
F9
F10
F11
F12
`
1
2
3
4
5
6
7
8
9
0
-
=
⌫
tab
Q
W
E
R
T
Y
U
I
O
P
[
]
\
caps lock
A
S
D
F
G
H
J
K
L
;
'
return
shift
Z
X
C
V
B
N
M
,
.
/
shift
fn
⌃
⌥
⌘
⌘
⌥
←
↑ ↓
→
Start recording

Everything you said,
you can find again.

What you said doesn't disappear — you can pull it back any time to copy, export, or re-transcribe. The things you used to keep in a clipboard manager now just live in your history.

  • Copy or export anything you've sent and reuse it any time
  • Not happy with the result? Re-transcribe the same recording
  • You decide how long to keep it: 7, 30, 90 days, or forever
  • iCloud sync is coming, but we never store your content
14:08Sprint review 我們聊了 vector store 的 latency⋯⋯0:23
13:42回 Slack：那個 PR 我晚點 review，先 ship⋯⋯0:08
13:21我剛剛用 Claude Code 試了一下 vertex，它真的順很多。0:12
11:55寄信給 Wendy：謝謝你昨天的 feedback，我們會在下版本⋯⋯1:04
Copy textExportRe-transcribe
Waitlist

Be the first to know when we launch.

Voiced is putting the finishing touches in place. Leave your email to join the waitlist, and we'll notify you the day we launch — get early access to voice input built for Traditional Chinese.

Voiced is still in beta. The interface and features shown here may differ from the public release.

A few answers first.

I've used dictation tools like Wispr Flow and Typeless — how is Voiced different?
Those tools are mostly English-first, with Traditional Chinese as an afterthought; they often output Simplified characters or mainland wording, so you end up proofreading every time. Voiced flips this: Traditional Chinese comes first, with native wording, Taiwanese terms, full-width punctuation, and Chinese-English code-switching all tuned by default — paste it and it's ready to use.
Where does my voice go? Who can see it?
Your Mac sends audio directly to the ASR provider, never through Voiced servers. Once it becomes text, that text briefly passes through Voiced servers on the way to the polish model. Voiced servers and providers use ZDR and no-training terms, releasing the content after processing instead of storing it.
Does Power Mode read the contents of the app I'm using?
Yes, when you turn on Power Mode or per-app style learning. Voiced knows which app you are using and uses Voiced transcripts and rewrites created in that app to summarize its polish style. It is off by default, and you can turn it off or delete history and dictionaries. Audio never passes through Voiced servers; polish text only passes through briefly under ZDR handling and is not stored.
Can I use it offline?
Not yet — it currently needs a connection. Cloud ASR and Polish are more reliable for Chinese-English code-switching and Traditional Chinese output, and that's the trade-off we make at this stage.
Does Voiced support iOS?
We're launching on macOS first (macOS 14+). A Windows version is planned; iOS is more constrained by the platform and will come later.
Does it sync with iCloud?
Everything is local by default. You can choose whether to sync your dictionaries and settings through iCloud Drive; history does not sync by default to protect your privacy.
Does it understand Cantonese?
Right now we prioritize Mandarin, Traditional Chinese output, and Chinese-English code-switching. Cantonese may work, but it isn't the main focus at this stage.

Just speak.

Voiced is designed for macOS and launching soon. Leave your email and we'll notify you the moment it's live.

Join the waitlist
macOS 14.0+ · Windows version in progress
Contact Us

Questions?
Leave us a message.

From product questions and waitlist notes to bug reports — we are glad to hear from you. For enterprise questions, team rollout, or business partnerships, you can reach us here too. Leave an email we can reply to, and we will get back to you soon.

Only used to reply to this message.