Next Line Mandarin — synthetic audio sources and limitations
Recorded license verification date: 2026-10-13

These MP3 files were synthesised locally from our original teaching text.
They are not copied lesson recordings, real calls, personal voice clones,
human-teacher recordings, or authentic WeChat voice messages.

Model: myshell-ai/MeloTTS-Chinese
Author: MyShell.ai
Model card: https://huggingface.co/myshell-ai/MeloTTS-Chinese
Upstream code: https://github.com/myshell-ai/MeloTTS
License declared by publisher: MIT (commercial use permitted).
The exact MIT notice shipped with the ONNX model is retained in MeloTTS-MIT.txt.

ONNX export: k2-fsa/sherpa-onnx, vits-melo-tts-zh_en
Source: https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/vits-melo-tts-zh_en.tar.bz2
Documentation: https://k2-fsa.github.io/sherpa/onnx/tts/pretrained_models/vits.html
Runtime: sherpa-onnx 1.13.8, Apache-2.0 (see sherpa-onnx-APACHE-2.0.txt).
The model and runtime are not distributed with the website.

Voice: ZH / speaker 0, a single synthetic voice for both practice-call roles.
Generation speed: 1.0. Short pauses are inserted between call turns.
Encoding: mono MP3, 24 kHz, 64 kbit/s; no music or third-party sound effects.
The optional slower playback changes browser speed to 0.8x, not the recording.

Rights scope:
This is a documented permissive-license workflow, NOT a declaration that
the model, training data, voice, or output is copyright-free or public domain.
The publisher's license does not independently prove all training-data or
personality rights. No copyright exclusivity or zero-risk guarantee is made.
Do not represent this audio as a real person, teacher endorsement or field recording.

Quality limits:
Synthetic speech may have pronunciation, rhythm or emphasis errors.
Technical playback/file/transcript checks are not independent teacher review.
Independent language and listening review and real learner testing are pending.
Text, dictionary-tone pinyin and English are provided beside selected clips.
Use the text when playback fails; adjust example details to your real situation.
Real calls may be faster, accented or noisy. These clips do not guarantee
comprehension of an actual call or agreement from another person.