Hear it. Follow its pitch.
The orange path follows the reference recording you hear. The green path follows your own microphone recording.
The reference voice
Our Mandarin curriculum uses computer-generated recordings made locally with MeloTTS by MyShell.ai and its contributors, using the MeloTTS Chinese voice model. Both are released under the MIT license. We credit their voice technology; AoA Dynamics provides the learning tools and pitch-practice interface.
Short words are generated with a brief lead-in to give the voice engine enough context, then extracted using its timing alignment. The recording you hear contains the target word. Playback volume is balanced across the curriculum.
Read the MeloTTS copyright and license notice
The audio-generation process also uses Google's multilingual BERT model, released under Apache 2.0. Read its license.
What the paths mean
Each reference path is estimated directly from the audio file played by Hear reference. It is not drawn from a tone number or a pinyin template. Gaps appear where there is no stable detected pitch, including silence and many consonants.
The vertical scale compares pitch relative to each speaker's middle range. The horizontal scale uses actual seconds. A deeper voice can follow the same relative movement as a higher voice. Different pacing moves the lines horizontally.
This is a listening and pitch-practice tool. It does not grade consonants, vowels, or whether you said the correct word. A synthesized voice is one reference example; natural speakers vary.
Written pinyin can show the dictionary tone while connected speech changes its realization. For example, two third tones together normally change the first syllable's spoken tone. Use the grammar and pronunciation lessons to understand those changes.
Your microphone recording stays in your browser. Bundled curriculum audio plays from this website; it does not call a speech service.
Return to Mandarin trainer