审阅范围 · Review scope
限定范围独立审阅。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。
Independent review within the stated scope. Keep the evidence and limitations stated in the text in view. This is not final corpus acceptance or validation of performance or learning outcomes.
先分清要听什么：和弦层次、放键与余响
本包通过限定独立来源文字／静帧、双语含义、初学者说明与符号审阅，未发现实质问题；出版时补清时长来源。T2是指定静态证据的作者归属，G是本例限定定义，O是本库原创解释或纸面设计。当前通过范围另列publication_review，来源逐项布尔值保留作者冻结时快照；不表示原音、动作、身体适合性或学习效果验收。
一个和弦的音都正确，却不知道该突出谁；两次起音相隔一样久，却不知道前一个键应保持多久。这是两个不同问题。先在纸上明确目标，再决定什么需要实际听、什么需要完整动作观察。
本包新读NC-084与NC-087全部289段既有ASR文字，检查17个选定静帧；其中NC-084的273秒画面模糊，不作精确主张依据。50段非教学／不确定候选排除出归属，不是听校确认的50处错误。910.635秒是继承的ASR记录时长合计；同一哈希的MP4容器元数据合计910.591秒，差异原因未查明。二者均不是观看或听校时长。没有新增ASR、原音听校、连续动作、完整谱例或学习者测试。
下面的小任务完全在纸上完成，不需要钢琴、唱歌、录音、MIDI或麦克风。音名C4指中央C，E4是其右边第二个白键；数字4是八度，不是手指号。88键编号包含黑键，以最左A0为1，因此C4=40、E4=44。编号用于核对位置，不要求实际伸手。
先会把一拍等分，并能区分“开始”和“结束”。不熟悉和弦也可先做时间任务；和弦小例只需认出C4、E4、G4是三个不同音，G4在最高处，不需要先弹三音。
纸面答案只证明某个符号关系算对了。未实际听就记“听感未验证”；未观察连续动作就记“动作未知”。两课都不能由ASR或静帧升级为完整视听审阅。
有边界的作者归属
B35-S01 · T2
Wiwi在同一和弦配置上比较不同突出位置，并使用“明亮”“宽阔”等个人听感标签。这支持提出“哪一层突出”的问题；本包没有听验这些标签，也不把最高音较响定为通用答案。
NC-084静帧173、194、359秒的明暗音符与文字；ASR第25–26、52–59、74–82段仅作上下文定位。未逐音复制或验收来源七音配置。
B35-S02 · T2
Wiwi用方块表示持键区间，另画余响，并比较分离、相接和重叠的区间。图能支持这些概念的区分；它不能单独证明某架琴的衰减秒数或某段演奏的清晰度。
NC-087静帧200、225、265、322秒；波形静帧114、140、157、166秒仅作未校准记录图观察。ASR第68–98段用于定位，完整来源曲例和原音未验。
和弦层次：音高身份与突出目标分栏
G：这里的“层次”指希望某些音更容易被注意，不是把音搬高。O：自编纸面例C4–E4–G4仍是这三个音，不论把G4记成前景还是暂不指定突出对象。先保持音高和起音位置，再写听感目标；实际能否听出目标另查。
O：给同一份C4、E4、G4事件写两条目标：A“让G4突出”，B“暂无突出指定”。它们不是两个和弦身份，也没有给出分贝或力度数值。B的“暂无指定”不等于已经测得三音等响。
G：piano、forte等力度标记给出音乐语境中的响度方向；它们不直接给每个和弦音一个精确比例。voicing可指选音／排列，也可指演奏中的音量层次，必须写清当前意思。声音听起来响、手上用力和MIDI velocity不是可以直接互换的一列数值。
O：拿到自己的合法谱面，先分列“印刷要求”“本人选择”“还不知道”。只有旋律／声部和上下文已确认，才把突出目标落到特定音符。没有明确前景时可保留未知，不能凭“最高”自动补出作曲家的意图。
O：先做纸面记录：固定这一小段的音与节奏，圈出拟突出线条，写一句“我要听到什么”。若之后实际弹奏，先分别确认音、节奏与舒适性，再比较同一短段的层次。记录“听出／未判定／未尝试”；不要用固定重量、手指高度或越用力越好的规则补空白。
O：休息后先遮住自己的标注，再从谱面重新说明选择及理由；次日可再做同例检查。最后把目标放回前后乐句，允许上下文改变选择。这是原创项目检查方法，没有证明某种层次最好或学习效果必然提高。
奏法：起音间隔、持键时间、余响是三栏
G：起音间隔是两个开始时刻之差；持键时间是某一个键的释放时刻减去开始时刻；余响是释放之后可能仍被记录或听到的声音。前两项可作为书面目标计算，第三项需要声音证据；不能从方块长度推出声音结束时刻。
G：articulation（奏法）涉及音的起止与连接方式。legato（连奏）和staccato（断奏）不能由本包三个纸面比值自动判定；谱面时值、实际放键与听到的延续不是同一概念。踏板、音区、乐器和空间是另须记录的条件，不由这里的算术控制。
O：若问题是“下一音晚了”，先查起音位置；若起音没变而连接不同，再分别查保持／释放目标和实际声音。不要为追求清楚而擅改谱面节奏，也不要从一幅波形图诊断手指动作。
两个起点不变，先预测键区间
只做纸面：原创4/4小节，以四分音符为一拍，四分音符=60 BPM，因此一拍=1秒。C4在offset 0开始，E4在offset 1开始；offset从0计，与从1开始的拍内计数不同。两音都记给RH（右手），LH（左手）无音，只是书面分工，不要求实际弹。先预测三版的起音间隔、各音持键秒数，以及前音释放到后音起点的有符号差。
差值=E4起点−C4释放点：正数是没有键保持的间隙，0是同一时点相接，负数的绝对值是两键保持区间重叠量。这个符号只描述键区间，不说明声音已停或听起来连／断。两区间同刻交接不算正时长重叠。
用原创事件表与等分时间条解释，不放来源谱图。时间条每格0.25拍；C/E表示只有相应键处于保持目标，X表示两键同时保持，·表示没有保持键。时间条是纸面目标图，不是录音波形或动作测量。
| 版本 | 事件 | 音高／键号 | 起点 | 释放 | 手 |
|---|---|---|---|---|---|
| A | A1 | C4 / 40 | 0 | 0.5 | RH |
| A | A2 | E4 / 44 | 1 | 1.5 | RH |
| B | B1 | C4 / 40 | 0 | 1 | RH |
| B | B2 | E4 / 44 | 1 | 2 | RH |
| C | C1 | C4 / 40 | 0 | 1.25 | RH |
| C | C2 | E4 / 44 | 1 | 2.25 | RH |
先预测，后核答案
答案：三版起音间隔都为1拍=1秒。A每音保持0.5秒，差值+0.5秒；B每音1秒，差值0；C每音1.25秒，差值−0.25秒，即键区间重叠0.25秒。A/B/C最后释放分别是offset 1.5/2/2.25；到小节结束4之间都没有保持键。没有一版提供余响时长或实测响度。
A: CC··EE··········
整只手没有保持键的拍坐标区间：LH=[[0, 4]]; RH=[[0.5, 1], [1.5, 4]].
B: CCCCEEEE········
整只手没有保持键的拍坐标区间：LH=[[0, 4]]; RH=[[2, 4]].
C: CCCCXEEEE·······
整只手没有保持键的拍坐标区间：LH=[[0, 4]]; RH=[[2.25, 4]].
若写出“C听起来一定连奏”，把结论改成“C的目标键区间重叠；听感未验证”。若把C的两次1.25相加说小节长2.5秒，回到时间轴：全小节仍4秒，两音重叠时不能把它们相加当经过时间。
把答案和时间条盖住，重新画完C；再核对最后释放与剩余空档。次日同例重做只叫保持检查，不叫未见材料迁移。把方法用到实际谱面时，先核原谱起音／时值／奏法，另写练习目标；不能直接把本例比值套成作品的标准奏法。
纸面完成证据是三版6个事件与全部手部空档核对一致，并正确保留“余响／演奏未知”。没有实际演奏证据就不填写演奏通过。这个小例不证明会读完整乐谱、会控制手指或已提高听觉能力。
剩余边界与相关阅读
两个来源的原声示范、连续动作、复杂和弦与Mozart片段均未完整验证。ASR中数值／专名疑点被隔离，不悄悄修成作者原话。原始文件、转录与静帧保留本地，本包只含短归属、原创解释和纸面事件。最终全库质量检查仍未完成。
本包来源范围
纸面关系检查
First define what to hear: chord balance, release and resonance
This package passed bounded independent source-text/still, bilingual-meaning, beginner-instruction and symbolic review with no material findings; publication clarifies duration provenance. T2 marks attribution from specified static evidence, G scoped definitions, and O original explanation or paper design. publication_review records the current passed scope while per-source booleans retain the author-time snapshot; this is not source-audio, motion, body-suitability or learning-efficacy acceptance.
A chord may contain the correct notes while its intended foreground remains unclear. Two attacks may be equally spaced while the earlier key’s holding time remains undecided. These are different questions. Define the target on paper first, then identify what requires listening or continuous motion observation.
This package newly reads all 289 existing ASR segments of NC-084/NC-087 and inspects 17 selected stills; NC-084’s blurred 273-second frame does not support precise claims. Fifty noninstructional/uncertain candidates are excluded from attributions, not 50 listening-confirmed errors. 910.635 seconds totals inherited ASR-record durations; metadata of the same-hash MP4 containers totals 910.591 seconds, with the cause of the difference unestablished. Neither value is viewing or audition time. No new ASR, original-audio audition, continuous-motion inspection, complete score-example review or learner test occurred.
The small task below is entirely on paper: no piano, singing, recording, MIDI or microphone is required. C4 is middle C; E4 is the second white key to its right. 4 denotes the octave, not a finger. 88-key numbering includes black keys, starting at leftmost A0=1, so C4=40 and E4=44. These numbers check location without requiring a physical reach.
First distinguish a start from an end and divide a beat equally. Begin the timing task even without chord knowledge. The chord example only requires recognizing C4/E4/G4 as three different pitches, with G4 highest; playing the three notes is unnecessary.
A paper answer establishes only the specified symbolic relationship. Without listening, record “listening result unverified”; without continuous observation, record “motion unknown.” Neither lesson can be upgraded from ASR/stills to complete audiovisual review.
Bounded author attributions
B35-S01 · T2
Wiwi compares different emphasis locations within one chord configuration and uses personal descriptions such as bright or broad. This supports asking which layer should stand out; we have not auditioned those labels and do not make a louder highest note the universal answer.
NC-084 stills at 173, 194, 359 seconds: differentiated note shades and captions; ASR segments 25–26, 52–59, 74–82 locate context only. The source’s seven-note configuration was neither reproduced nor accepted note by note.
B35-S02 · T2
Wiwi depicts key-holding intervals as blocks, draws separate resonance tails, and compares separated, adjoining and overlapping intervals. These drawings support the conceptual distinction, but alone establish neither a piano’s decay time nor a passage’s clarity.
NC-087 stills at 200, 225, 265, 322 seconds; waveform stills 114, 140, 157, 166 are observations of uncalibrated recording plots only. ASR segments 68–98 locate context; complete source score examples and original audio remain unverified.
Chord balance: separate pitch identity from intended foreground
G: balance here means intending some tones to attract more attention, not moving them higher. O: our paper example C4–E4–G4 keeps those three pitches whether G4 is foregrounded or no foreground is specified yet. Preserve pitches and attack positions before stating the listening goal; whether the goal is audible is a separate check.
O: attach two goals to the same C4/E4/G4 events: A: “foreground G4”; B: “no foreground specified yet.” These are not two chord identities and contain no decibel or attack-strength values. B’s unspecified goal does not establish measured equal loudness.
G: dynamic markings such as piano and forte give contextual loudness directions; they do not directly assign an exact ratio to every chord tone. Voicing can concern pitch selection/arrangement or performed balance, so specify the intended meaning. Perceived loudness, bodily force and MIDI velocity are not interchangeable numerical columns.
O: with your lawfully obtained score, separate printed requirements, your choices and unknowns. Assign foreground to particular notes only after checking voice/melody and context. Leave it unknown when unclear rather than inferring the composer’s intention merely from the highest pitch.
O: first record on paper: keep the passage’s notes/rhythm, circle the proposed foreground and write what you want to hear. If you later play it, first check pitches, rhythm and comfort separately, then compare balance in the same short passage. Record heard/undetermined/not attempted; do not fill gaps with fixed weight, finger-height or greater-force rules.
O: after a break, hide your annotation and explain the choice and reason from the score again; repeat on the same example the next day if useful. Then return the target to its surrounding phrase, allowing context to change the choice. This is our project-check design, not proof of an optimal balance or guaranteed learning improvement.
Articulation: attack spacing, holding time and resonance are three columns
G: attack spacing is the difference between two starting times; a key-hold duration is its release time minus its starting time; resonance is sound that may still be recorded or heard after release. The first two can be calculated as written targets; the third needs sound evidence. Block length cannot determine when sound ends.
G: articulation concerns how notes begin, end and connect. Legato and staccato cannot be assigned automatically from this package’s three paper ratios. Notated duration, actual key release and perceived continuation are different concepts. Pedal, register, instrument and space are conditions to record separately, not variables controlled by this arithmetic.
O: if the question is whether the next note was late, check attack positions first. If attacks are unchanged but connection differs, examine hold/release targets and actual sound separately. Do not alter written rhythm merely to seek clarity, or diagnose finger motion from a waveform.
Keep two starts fixed and predict key intervals
Paper only: an original 4/4 bar uses quarter-note beats at quarter note=60 BPM, so one beat=1 second. C4 starts at offset 0 and E4 at offset 1; offsets begin at 0 rather than the beat count’s starting value of 1. Both notes are assigned to RH and LH has no notes; this is written allocation without a playing requirement. First predict each version’s attack spacing, each hold in seconds, and the signed difference from the earlier release to the later attack.
Difference=E4 start−C4 release: positive means a gap with neither key held, 0 means adjoining at one time point, and the absolute value of a negative difference is the overlap of the two holding intervals. This describes key intervals only, not acoustic silence or audible legato/staccato. A shared boundary alone is not positive-duration overlap.
Use original event tables and equal-cell timelines, without source score images. Each cell is 0.25 beat; C/E means only that key is targeted as held, X means both held, and · means neither held. The timeline is a paper target, not a recording waveform or motion measurement.
| Case | Event | Pitch / key | Start | Release | Hand |
|---|---|---|---|---|---|
| A | A1 | C4 / 40 | 0 | 0.5 | RH |
| A | A2 | E4 / 44 | 1 | 1.5 | RH |
| B | B1 | C4 / 40 | 0 | 1 | RH |
| B | B2 | E4 / 44 | 1 | 2 | RH |
| C | C1 | C4 / 40 | 0 | 1.25 | RH |
| C | C2 | E4 / 44 | 1 | 2.25 | RH |
Predict first; answers follow
Answer: every version has attack spacing 1 beat=1 second. A holds each note 0.5 second with difference +0.5 second; B holds each 1 second with difference 0; C holds each 1.25 seconds with difference −0.25 second, meaning 0.25 second of key-interval overlap. A/B/C release their last note at offset 1.5/2/2.25 respectively, with neither key held until the bar ends at 4. No version supplies resonance duration or measured loudness.
A: CC··EE··········
Whole-hand gaps in beat offsets: LH=[[0, 4]]; RH=[[0.5, 1], [1.5, 4]].
B: CCCCEEEE········
Whole-hand gaps in beat offsets: LH=[[0, 4]]; RH=[[2, 4]].
C: CCCCXEEEE·······
Whole-hand gaps in beat offsets: LH=[[0, 4]]; RH=[[2.25, 4]].
If you wrote “C must sound legato,” replace it with “C’s target key intervals overlap; listening result unverified.” If you added C’s two 1.25 durations and called the bar 2.5 seconds, return to the timeline: the whole bar remains 4 seconds; simultaneous holds do not sum to elapsed time.
Hide the answers and timelines, redraw C completely, then check its final release and remaining gap. Repeating this disclosed example the next day is a retention check, not unseen-material transfer. For a real score, first verify its attacks, durations and articulation, then write a separate practice target; these ratios are not a work’s standard articulation.
Paper completion means matching all 6 events and every hand-level gap across the three cases while retaining “resonance/performance unknown.” Without evidence of actual performance, do not mark performance passed. This example does not establish complete-score reading, finger control or improved hearing ability.
Remaining limits and related reading
Neither source’s sound demonstrations, continuous motion, complex chords or Mozart excerpt has been fully verified. Suspect ASR numbers/names are quarantined rather than silently rewritten as the author’s words. Originals, transcripts and stills remain local; this package contains only brief attributions, original explanations and paper events. Final whole-corpus review remains incomplete.
Sound targets and motion evidence
Separate tempo from note duration
This package’s source scope
Paper-relationship checks