Description
Taking notes in a meeting, keeping up in a chat, mixing Cantonese and English in one sentence: classic input methods often get the speech wrong and leave you fixing characters one by one. Doubao IME hands both voice recognition and prediction to a large model: it understands dialects, technical terms and mixed Chinese and English speech, and when typing it predicts long phrases and fixes slips as you go, using the same speech recognition as the Doubao assistant.
It does more than turn sound into text. Loose spoken phrasing is tidied into fluent writing, with unfinished thoughts filled in, and text copied on a phone can be pasted straight on the computer without a chat app in between. The official site offers Windows, macOS, iOS, Android and HarmonyOS builds; the Windows version is currently V0.9.2 and free to use. The interface is Chinese only.
Voice input: hold the voice key to speak and release to finish, with dialects, technical terms and mixed Chinese and English supported, and a button to undo the input
Noisy environments: tuned for real places such as a subway or an office, and stays stable in noise or when you whisper
LLM typing: smart prediction, long phrase completion and live correction, so even short pinyin or typos can still return the sentence you meant
Smart text cleanup: turns scattered speech into connected text, completes unfinished thoughts and picks out the key points
Cross device paste: copy on the phone and paste on the computer directly, with no relay step in between
Many platforms: versions for Windows, macOS, iOS, Android and HarmonyOS NEXT
It does more than turn sound into text. Loose spoken phrasing is tidied into fluent writing, with unfinished thoughts filled in, and text copied on a phone can be pasted straight on the computer without a chat app in between. The official site offers Windows, macOS, iOS, Android and HarmonyOS builds; the Windows version is currently V0.9.2 and free to use. The interface is Chinese only.
Features
Voice input: hold the voice key to speak and release to finish, with dialects, technical terms and mixed Chinese and English supported, and a button to undo the input
Noisy environments: tuned for real places such as a subway or an office, and stays stable in noise or when you whisper
LLM typing: smart prediction, long phrase completion and live correction, so even short pinyin or typos can still return the sentence you meant
Smart text cleanup: turns scattered speech into connected text, completes unfinished thoughts and picks out the key points
Cross device paste: copy on the phone and paste on the computer directly, with no relay step in between
Many platforms: versions for Windows, macOS, iOS, Android and HarmonyOS NEXT
