审阅范围 · Review scope
原创合成视频 · 数值与抽帧检查。请保留正文中的出处与限制；不代表全库最终验收、实际演奏或学习效果验证。
Original synthetic video · numerical and selected-frame checks. Keep the evidence and limitations stated in the text in view. This is not final corpus acceptance or validation of performance or learning outcomes.
音符更多，拍速相同
这是原创教学预览：用4/4、四分音符=60 BPM说明音符密度与拍速的区别。先轻点稳定的拍，再看一拍被分成几个相等部分。点击视频，或下载后离线播放。音符用440 Hz合成音，拍点用1760 Hz短音；不是钢琴录音或真实动作示范。旁白是英文合成语音，画面同时有完整中文和英文字幕，没有中文配音。
| 播放范围（秒） | 每小节起音数 | 相邻起音间隔 | 可核对的问题 |
|---|---|---|---|
| 13–17 | 4 | 1秒 | 一拍一个四分音符 |
| 21–25 | 8 | 0.5秒 | 一拍两个八分音符 |
| 29–33 | 16 | 0.25秒 | 一拍四个十六分音符 |
三个示范的每小节均为4秒。图中格子表示分配给音符的时槽，不声称合成音的包络恰好填满时槽。声学振动的Hz、音乐拍速的BPM和每秒起音数不是同一个量。
原创自查：先暂停在13秒前，预测后两段的拍点是否会变密；再播放，分别数拍点和音符。答案：拍点始终每秒一个，后两段只增加拍内起音。若数乱，先只点拍，随后单独数细分；无需上琴或固定指法。下一次重看前先口述规则，这是自查而非经过验证的训练剂量。参见节奏主题与拍速互动。
字幕、源文件与验证
中文字幕VTT／SRT，英文字幕VTT／SRT。可编辑源包含原创画面程序、已生成的英文旁白片段及其中文文字对应、字幕、字体和许可。修改图形与重新合成可用已有片段，不需要下载模型或启动语音生成。源包内有重建说明。
数值核查记录45.033秒、1920×1080、60 fps及2702帧；本次重建音频4条结果逐字节匹配原音频，并重新检测4/8/16起音、1秒拍距、无旁白的示范段和56个编码边界帧。另人工看过3张本次解码代表帧。视频文件是原交付文件，未用适配后的源包重新编码全片。
仍未验证：人耳试听、实体iPad播放、学习成效。时间峰值包括合成包络及滤波延迟，不应当成精确起音偏差。原作者已看过13张教学帧，独立父任务看过3张；这些与本次检查分别记录，不相加为连续观看。预览不代表全库最终验收。
原创程序和合成声用于此教学样例；辅助代码保留Clean Video的MIT许可，Inter和Noto Sans SC字体保留OFL许可文本。旁白使用已有Kokoro af_heart生成片段，模型卡声明权重为Apache-2.0；这不代表对训练数据权利进行了独立审计。没有原课程媒体。二进制文件哈希见清单。
More notes, the same tempo
This original teaching preview uses4/4 with quarter note=60 BPM to distinguish note density from tempo. Tap the steady beat, then observe how many equal parts divide each beat. Open the video or download it for offline playback. Notes use440 Hz synthesized tones and beat markers use1760 Hz short tones; these are not piano recordings or real movement demonstrations. Narration is synthetic English with complete Chinese and English open captions, without Mandarin narration.
| Playback range (seconds) | Attacks per bar | Interval between attacks | Check |
|---|---|---|---|
| 13–17 | 4 | 1 second | One quarter per beat |
| 21–25 | 8 | 0.5 seconds | Two eighths per beat |
| 29–33 | 16 | 0.25 seconds | Four sixteenths per beat |
Every demonstration bar lasts4 seconds. Cells represent allocated note slots, not a claim that each synthesized envelope fills its slot exactly. Acoustic Hz, musical BPM and attacks per second measure different things.
Original self-check: pause before13 seconds and predict whether beat markers will become denser in the next two demonstrations. Play and count beats separately from notes. Answer: beat markers remain one per second; only attacks within beats increase. If counting becomes confused, tap beats alone, then count subdivisions separately; no keyboard or prescribed fingering is needed. Explain the rule before watching next time. This is a self-check, not a validated training dose. See the rhythm topic and tempo interaction.
Captions, source and verification
Chinese VTT/SRT, English VTT/SRT. The editable source contains original rendering code, already generated English narration clips with Chinese text counterparts, captions, fonts and licenses. Existing clips support graphic edits and reassembly without model downloads or speech generation. Rebuild instructions are inside.
Numerical checks record45.033 seconds,1920×1080,60 fps and2702 frames. This integration rebuilt4 audio outputs that match the originals byte for byte, and freshly detected4/8/16 attacks,1-second beat spacing, narration-free demonstrations and56 encoded boundary frames. Three newly decoded representative frames were also visually inspected. The video bytes are the original delivery; the adapted source package has not re-encoded the entire movie.
Unverified: human listening, physical iPad playback and learning efficacy. Measured peak times include envelope/filter delay and are not exact onset errors. The producer inspected13 teaching frames and the parent independently inspected3; these are recorded separately from this integration and do not add up to continuous viewing. This preview does not establish final corpus acceptance.
Original code and synthesized sound serve this illustration. Helper code retains Clean Video's MIT license; Inter and Noto Sans SC retain OFL texts. Existing narration clips used Kokoro af_heart; its model card declares Apache-2.0 weights, without implying an independent training-data rights audit. No purchased-course media are included. Binary hashes are in the manifest.