Mixed-language dictation
Speak Chinese and English
in the same sentence.
Bilingual people rarely stay in one language. TypeAloud's local model detects what you are speaking as you go, so a sentence that starts in Chinese and ends in English is written down as you said it.
No language switch to remember
Most dictation tools ask you to pick a language first, then mangle anything outside it. TypeAloud's default local model, Qwen3-ASR, takes no language setting at all. It works out the language from the audio.
That fits how technical work in Chinese actually sounds: product names, API names and half the nouns are English, and the grammar around them is Chinese.
Built for long dictation
Hold Ctrl+Shift for as long as you are talking, then release. Local mode shows progress as text arrives. Completed transcripts go to History, and a failed transcription can be retried when its audio was saved, so a long recording is not lost to one error.
Translate into English when you need to
Hold Shift+Alt instead and TypeAloud writes what you said in English. Translation runs through a cloud provider with your own API key, so it is optional and separate from the offline local mode.
Free, on your device
Local transcription needs no account, subscription or API key, and has no word limit. Audio is processed on your Mac. Apple Silicon is recommended.
Common questions
- Do I have to set Chinese or English before speaking?
- No. The local Qwen3-ASR model detects the spoken language automatically.
- Does it handle English terms inside Chinese sentences?
- Yes, that is the case it is designed for. Accuracy on unusual names still varies, as with any speech model.
- Is the interface available in Chinese?
- Yes. The interface language is a separate setting from the language you speak.
- Does it work offline?
- Local transcription works offline after a one-time model download of about 1 GB. Translation needs a cloud provider.
Local transcription. No account. No subscription.
Download for Mac