Little Motif 小小乐句
INTERACTIVE WIREFRAME · 交互线框Connecting…

A listening adventure / 一次聆听冒险

Where can your little tune go?

你哼出的旋律,会走到哪里?

Make a little tune, watch it change, then discover a real piece that sounds a bit like yours.
哼一段、听它变化,再找一首与你的旋律相呼应的真实作品。

01

Make your little tune

哼一小段,或先选一个样例试试。

Try 3 to 8 seconds of humming · 哼唱 3–8 秒

Your microphone stays on this device. / 麦克风音频仅留在此设备。

Or try an example 或试试样例

The examples are starting tunes only. Your created piece still comes from the AI service.

02

Is this your tune?

先听听它是不是你哼出的旋律。

Choose an example or hum a tune to see its shape here.
选择样例或哼唱后,这里会出现旋律的形状。
03

Give it a musical journey

选一种感觉,让自己的主题在音乐里出发、变化、回来。

A short AI score; sound is played by your browser. / AI 创作短乐谱,浏览器合成声音。
04

Hear an echo in classical music

听听真实作品里的主题,哪里像你的旋律?

Create a piece to unlock two listening trails.
创作完成后,可以找到两条聆听线索。
For grown-ups: what this wireframe does / 给成人:原型实现范围

A local SwiftF0 neural model listens for one pitch line and confidence, then the app groups stable notes and pauses. It can also follow pitched background sounds; child-vocal robustness has not been established. Only note numbers and durations go to OpenRouter for a new symbolic score, stored by the local server. Web Audio plays the score and drives its event map. An optional Lyria call makes a separate orchestral interpretation, whose notes and timing are not verified against that map. Comparison previews are synthesized excerpts from four curated classical themes. This is an adult-review prototype; no child learning study has been run.

浏览器使用本地 SwiftF0 神经音高模型,保留连续音高与置信度,再组成乐句。模型也可能跟到背景音乐;真实儿童哼唱的可靠性尚未建立。仅将音高与时值发送给 OpenRouter 创作乐谱,并由本地服务器保存。Web Audio 播放乐谱并驱动事件图;可选 Lyria 调用生成独立管弦演绎,其音符与图谱时间尚未核验。对听片段来自四个整理过的古典主题,使用合成试听。本原型用于成人评审,尚无儿童学习测试。