A Mandarin Chinese shadowing app: practise speaking by shadowing real Chinese videos, with feedback on every
tone and sound.
Runs on your own computer. Free and open source.
Download for Windows · User guide · Run from source
Paste a YouTube link, and a few minutes later the video is a lesson: cut into short sentences, with pinyin, every character aligned to the audio. Then you listen and repeat, sentence by sentence or a whole passage at a time, and the app tells you which tones and sounds you got right.
It's for learners who want to work on their pronunciation, tones and speaking, from beginner (HSK 1) up. Shadowing (跟读, or 影子跟读) means repeating a native speaker right after them; this app checks every syllable you say.
- Make lessons from YouTube or your own files. Chinese subtitles are used when the video has real ones; otherwise the speech recogniser writes the text.
- Practise two ways. Echo: one sentence at a time, listen then repeat. Flow: shadow a whole passage, after the speaker or along with them.
- See what went wrong. Every syllable gets a tone verdict (
3→2shows what you said instead), wrong initials and finals are pointed out, and pitch lines show your melody against the speaker's. - Hear yourself back. Every take is kept; play the original and your take back to back.
- Train your weak spots. Stats show your weakest tones, tone pairs, sounds and characters, and Drills build a short session from all your lessons around one of them.
- Review with Anki. Practised sentences become listening cards in one click.
- Look words up. The text works with the Yomitan pop-up dictionary.
| Echo: listen, then repeat | Flow: your turn |
|---|---|
 |
 |
| Stats: your weakest tones, sounds and characters | Library: lessons grouped by channel |
 |
 |
More screenshots
 |
 |
 |
 |
 |
 |
Screenshots show lessons made from Mandarin Click's HSK1 stories.
- Download
chinese-shadow-speech-<version>-setup.exefrom the latest release. - Run it. The installer isn't code-signed yet, so Windows may say "Windows protected your PC": click More info, then Run anyway. It installs for your user only, without administrator rights.
- The first start sets up the app and downloads the speech models (about 5 GB), which takes 10 to 20 minutes. A page in your browser shows the progress. Later starts take a few seconds.
- The Library opens. Paste a YouTube link to make your first lesson.
Needs: Windows 10 or 11 (64-bit), 8 GB of RAM, 8 GB of free disk space, and an internet connection for the first start and for YouTube. No graphics card needed.
The app keeps running in the notification area after you close the browser tab; use its icon to open it again or to quit. Updates, your data and uninstalling: see Your data on Windows.
On Linux, or to work on the code, see Run from source.
Everything runs on your computer: the speech models, your recordings and your scores never leave it. The app goes online only to download videos you ask for, to download the speech models once, to keep the YouTube downloader (yt-dlp) up to date, and to check GitHub for a newer version.
- Next: testing on more laptops, including long videos; lessons from Bilibili links and podcast feeds.
- Later: a Microsoft Store version; a signed installer.
Ideas and bug reports are welcome as issues.
- User guide: every page and feature, keyboard shortcuts, Anki setup, your data on Windows.
- Development: run from source, tests, the Windows package, where things are in the code.
- Configuration:
config.toml, CPU-only use, Docker, YouTube cookies and yt-dlp.
Free software under the GNU Affero General Public License v3.0 or later: you may use, study, share and change it; if you distribute a changed version, or run one as a service for others, share its source under the same licence. The speech models it downloads have their own licences (THIRD_PARTY.md); the alignment model is non-commercial (CC-BY-NC 4.0), so the app is for non-commercial use.