审阅范围 · Review scope
公开来源文字与原创综合已审阅。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。
Public-source text and original synthesis reviewed. Keep the evidence and limitations stated in the text in view. This is not final corpus acceptance or validation of performance or learning outcomes.
公開資料第十六批：創作、伴奏與編配判讀
Public batch 16: creation, accompaniment, and arranging interpretation
六篇公開創作與編配正文；為原聲鋼琴初學者設計中英雙語紙上任務及可選的極短演奏。 Six public creation/arranging article bodies, with bilingual paper tasks and optional very short playing examples for an acoustic-piano beginner.
完成：6/6 篇完整公開正文；媒體、軟體、原始工程審查：0。
Completed: 6/6 complete public bodies; media, software, and source-project reviews: 0.
閱讀與整合規則 / Reading and integration rules
原始中文題名、網址與來源 ID 完整保留；英文題名為編輯翻譯。JSON 與 Markdown 是同一成果的兩種表示，不重複計數。
Original Chinese titles, URLs, and source IDs are retained exactly; English titles are editorial translations. JSON and Markdown are two representations of one result, not double-counted reviews.
六篇均有完整公開正文；有會員提問或課程連結，不等於讀過會員作品或課程。僅審文字；內嵌音訊、影片、圖片、譜圖、軟體及原始工程未檢查。
All six have complete public article bodies. Member questions and course links do not establish review of member work or courses. Text alone was reviewed; embedded audio, videos, images, score images, software, and source projects were not inspected.
來源為作者自己的教學或製作說明；技術、聽感和成效的全部主張不因此成為實驗證據。產品、AI、價格、可用性與推薦不沿用為現今事實。
Sources are the author’s own exposition or production accounts, not experimental support for every technical, perceptual, or learning claim. Product, AI, price, availability, and recommendation statements are not carried forward as current facts.
來源衍生預算含雙語題名、摘要、用途、證據類型、涵蓋與限制，以每個中日韓字符加英數詞元保守合計，單篇不超過 200；原創任務獨立標示。
Source-derived budgets include bilingual titles, summaries, roles, evidence type, coverage, and caveats, conservatively counting each CJK character plus Latin/alphanumeric token, at most 200 per source. Original tasks are marked separately.
不複製歌詞、來源練習、旋律或譜圖；不補造沒有看見的聲部或配器。任務所有音符另行設計，音名表不證明真實樂器。
No lyrics, source exercises, melodies, or score images are copied; unseen parts and instrumentation are not fabricated. All task notes are newly designed, and pitch tables do not prove real instruments.
原聲鋼琴不必有任何電子功能；全部可先紙上完成，不需麥克風、錄音、軟體、帳戶或購物。未安裝、上傳來源、購買、登入、推送或授權麥克風。
The acoustic piano needs no electronic features. All tasks can begin on paper without a microphone, recording, software, account, or purchase. Nothing was installed, source-uploaded, bought, logged into, pushed, or granted microphone access.
先寫預測，再看公開答案。答案與變題均已披露，不能稱為未見測驗或已證實遷移；本次沒有使用者表現資料。
Write predictions before reading disclosed answers. Answers and variants are already public, so they are not unseen tests or demonstrated transfer; no user-performance data were collected.
原創共用流程 / Original shared workflow
每次只選一張任務卡，先讀假設與音高表；缺旋律、和弦、拍位、持續或手部分工時，不自己填成來源事實。
Choose one task card at a time and read assumptions and pitches first. Missing melody, chords, beat positions, durations, or hand assignments are not filled in as source facts.
分欄記「題給事實」「我的預測」「紙上核對」「可選試彈的感受」「尚未知」。不把各欄互相冒充。
Use separate columns for supplied facts, prediction, paper check, optional playing impressions, and remaining unknowns. Do not substitute one type for another.
可選試彈採舒適速度與音量，不訂重複次數或成果期限；分手、停頓、琶音化都可以，但記錄改了什麼。不適即停。
Optional playing uses comfortable pace and volume without repetition quotas or deadlines. Separate hands, pauses, and arpeggiation are allowed if the changes are recorded. Stop if uncomfortable.
完成標準是解釋資料與改動，而非猜作者偏好、買器材或聲稱進步。要說明學習遷移，需要日後另設未見題，這裡沒有做。
Completion means explaining data and edits, not guessing the author’s preferences, buying equipment, or claiming progress. Learning transfer would require later unseen tasks, which are not conducted here.
跨批預算 / Cross-batch budgets
未取用任何補充來源正文；以網址搜尋既有 public-batch JSON，未見這六篇的先前審閱摘要。索引與指派清單不算已讀正文，也不把鏡像檔重複累計。 No supplementary source bodies were used. URL searches of earlier public-batch JSON found no prior reviewed summaries of these six articles. Index/assignment entries are not body reviews, and mirror files are not double-counted.
1. 11 個給創作者的秘訣
11 tips for creators
WEB-9caf569afb5b · 原文 / Source
頁面日期 / Page date: 2022-02-06 · 標示篇幅 / Labeled length: 2231 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 30–113；擷取總行 / extraction total: 118.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
作者建議改變起點、限制選項，先留素材再修改，並用假想委託與期限推動完成。 The author suggests changing starting points, narrowing choices, preserving material before editing, and using fictional briefs and deadlines to finish.
用途 / Role
建立修改紀錄。 Keep revision records.
來源限制與編輯保留 / Source limitations and editorial qualifications
個人方法，不保證原創或效率。 Personal methods, not guarantees of originality or efficiency.
來源衍生預算 / Source-derived budget: 119/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: 無 / none.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
只改一格，保留兩版 Change one cell; preserve both versions
分清「有一個不同版本」與「證明作品更好」，並保留可追溯的修改理由。 Distinguish having a different version from proving a better composition, while preserving a traceable revision reason.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 紙筆即可；可選用原聲鋼琴。4/4，每小節四拍；C4 是中央 C。右手單音、左手全休、無踏板。每音一拍，R 代表休止一拍。 Paper is sufficient; acoustic piano is optional. In 4/4 each bar has four beats; C4 is middle C. Right-hand single notes, left hand resting throughout, no pedal. Every note lasts one beat; R means one beat of rest.
草稿 A / Draft A1.小節 / Bar 1
右手各拍 / Right-hand beats["G4", "A4", "F4", "R"]
2.小節 / Bar 2
右手各拍 / Right-hand beats["F4", "G4", "A4", "F4"]
修改契約 / Revision contract 只改第二小節第四拍：可選 G4 或 R，其餘七格不變。目標只定為「讓結尾事件與原版不同」，不是指定情緒。 Change only bar 2 beat 4 to G4 or R; leave the other seven cells unchanged. The sole objective is a different final event, not a prescribed emotion.
紀錄欄 / Log columns 預測｜實際選擇｜哪些格沒變｜選擇理由｜尚未證明什麼 Prediction | chosen edit | unchanged cells | reason | what remains unproven
先預測與作答 / Predict and try first
在看答案前，圈出唯一允許改動的位置，預測兩個選項各會改變音高還是音符／休止的種類。
Before the answer key, circle the sole editable location and predict whether each option changes pitch or the note/rest category.
先抄好 A，另寫 B；不要擦掉 A。選一項並寫一句理由；沒有琴也可完成所有結構判讀。
Copy A, then write B separately without erasing A. Choose an option and give one reason; every structural check can be completed without an instrument.
若要試彈，先逐格找鍵，再以舒適速度分別彈 A/B；可停下，不訂速度或成功次數。記自己的感受，也允許「沒有明顯差別」。
If playing, locate keys cell by cell, then play A/B separately at a comfortable pace. Pauses are allowed; there is no speed or repetition target. Record your own impression, including “no clear difference.”
已公開答案與推理 / Disclosed answers and reasoning
選 G4：只有最後一音由 F4 變 G4，兩版都七個音符事件。選 R：最後一拍由音符變休止，B 只有六個音符事件。兩版均保留八拍。
Choosing G4 changes only the final pitch from F4 to G4; both versions have seven note events. Choosing R replaces the final note with rest, leaving six note events in B. Both versions retain eight beats.
完成修改契約不代表更好聽、更原創或已學會遷移。可驗證的是修改位置、事件數及理由紀錄。
Meeting the revision contract does not prove better sound, greater originality, or transfer. The edit location, event count, and recorded rationale are checkable.
任務界線 / Task limits
這是特定兩小節的版本管理，不測創造力。原聲琴的殘響不會因記號 R 而保證瞬間消失。 This is version control for two specified bars, not a creativity test. An R symbol does not guarantee instantaneous disappearance of an acoustic piano’s resonance.
已公開變題 / Disclosed variant
已公開變題：改為只准改第一小節第二拍 A4→G4。音符事件數不變，修改座標不同；先寫預測再核對。這不是未見過的測驗。 Disclosed variant: permit only bar 1 beat 2, A4→G4. Note-event count stays unchanged while the edit coordinate changes. Predict before checking; this is not an unseen test.
完成檢核 / Completion check
能指出唯一改動、數對事件、保留 A/B，並分開結構結果與個人偏好。 Identify the sole edit, count events correctly, retain A/B, and separate structural results from personal preference.
2. Q+A | 用 AI 輔助音樂學習，寫出不套公式的鋼琴伴奏，以及如何練習即興
Q+A: AI-assisted learning, nonformulaic piano accompaniment, and improvisation
WEB-34ecf3e2a7f5 · 原文 / Source
頁面日期 / Page date: 2025-04-20 · 標示篇幅 / Labeled length: 4057 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 30–159；擷取總行 / extraction total: 179.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
創作時把 AI 當素材而非成品；重視伴奏音域與手部可行性。即興分構想與執行，技巧練習針對具體問題。 For creative work, treats AI as material rather than finished work; emphasizes accompaniment register and playability. Separates improvisational ideas from execution, with technique practice targeting specific problems.
用途 / Role
核對提案。 Check proposals.
來源限制與編輯保留 / Source limitations and editorial qualifications
2025年觀點，非今日AI評測；會員作品未讀。 2025 views, not current AI testing; member submissions unread.
來源衍生預算 / Source-derived budget: 183/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: original_feedback_app_acceptance_design.acceptance_cases[1].given.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
先核對，再選伴奏節奏 Verify first, then choose accompaniment rhythm
不用任何 AI 帳戶，練習核對一張假想建議，區分可計算錯誤、編配選擇與尚未測試的手部可行性。 Without an AI account, audit a fictional suggestion and distinguish checkable errors, arranging choices, and untested hand feasibility.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 紙上先做；可選原聲鋼琴。4/4，C4＝中央 C，無踏板；右手每音一拍，左手根音持四拍，兩手第一拍同時開始。可分手、可在小節間停下。 Start on paper; acoustic piano is optional. In 4/4, C4 = middle C, with no pedal. Each right-hand note lasts one beat; each left-hand root lasts four beats. Hands start together on beat 1. Separate-hand work and pauses between bars are allowed.
給定和弦拼寫 / Supplied chord spellings 本題給定 C 大三和弦的音名集合 {C,E,G}，F 大三和弦 {F,A,C}；不需採譜或猜和弦。 The task supplies C-major pitch names {C,E,G} and F-major pitch names {F,A,C}; no transcription or chord guessing is required.
版本 A / Version A1.小節 / Bar 1
和弦 / Chord C
左手 / Left hand音高 / Pitch C3
拍位 / Beat 1
時值（拍）/ Duration (beats) 4
右手各拍 / Right-hand beats["G4", "E4", "G4", "E4"]
2.小節 / Bar 2
和弦 / Chord F
左手 / Left hand音高 / Pitch F3
拍位 / Beat 1
時值（拍）/ Duration (beats) 4
右手各拍 / Right-hand beats["A4", "F4", "A4", "F4"]
虛構建議 / Fictional advice 完全虛構的建議卡：「第二小節的三和弦音名是 F、A、C♯。把左手四拍長音換成每拍一次 F3，就保證更有音樂性，而且人人不必試就能彈。」不是任何實際模型輸出。 Entirely fictional advice card: “Bar 2’s triad is F, A, C♯. Replacing the four-beat left-hand note with F3 on every beat guarantees more musicality, and everyone can play it without trying.” This is not an actual model output.
版本 B 契約 / Version B contract 只把第二小節左手改為第一至第四拍各彈一次 F3，每次一拍；第一小節與全部旋律維持 A。 Change only bar 2’s left hand to F3 on beats 1–4, each lasting one beat; preserve bar 1 and the entire melody from A.
先預測與作答 / Predict and try first
先把建議分成三句，逐句標「與給定資料矛盾」「可試的版本」「尚未有證據」。不要用講得自信當成答案正確。
Split the advice into three claims and label each “contradicts supplied data,” “testable version,” or “not evidenced.” Confidence of wording is not correctness.
在看答案前，數 A/B 左手音符事件；預測 B 改的是節奏、音高、旋律中的哪幾項。
Before the key, count left-hand note events in A/B; predict which of rhythm, pitch, and melody B changes.
可選：先單手找 C3 與 F3；再試第一小節，最後試第二小節 A/B。若合手暫不方便，分手比較並明記「未合手」，不冒充完整伴奏測試。
Optional: locate C3/F3 with one hand, try bar 1, then bar 2 in A/B. If coordination is not yet comfortable, compare hands separately and record “not combined,” rather than claiming a full accompaniment trial.
自己再選一次：保留長音，或採用已寫明的重複音。理由只寫可觀察事項或個人偏好，不要求選較密的版本。
Choose again for yourself: retain the held note or use the written repetitions. State an observable reason or personal preference; denser is not required.
已公開答案與推理 / Disclosed answers and reasoning
C♯ 不在題給的 {F,A,C}，應改 C。這只核對本題，不評比現今 AI 的能力。
C♯ is absent from the supplied {F,A,C}; replace it with C. This checks the task, not today’s AI capabilities.
A 左手總共兩次起音；B 共五次。第二小節由一次四拍變四次一拍，根音 F3 與右手旋律未變。
A has two left-hand attacks in total; B has five. Bar 2 changes from one four-beat note to four one-beat notes; root F3 and the right-hand melody are unchanged.
重複音版本可以試，但保證更好聽與人人都可彈都未成立。未合手的結果不能充當合手能力證據。
The repeated-note version is testable, but guaranteed musical improvement and universal playability are unsupported. Separate-hand results do not establish hands-together ability.
任務界線 / Task limits
這八拍不涵蓋整首歌、所有伴奏或即興能力。各次演奏未錄音，不能當嚴格控制的聽感實驗；有不適即停止。 These eight beats do not represent a whole song, all accompaniment, or improvisational ability. Unrecorded retakes are not a strictly controlled perception experiment; stop if uncomfortable.
已公開變題 / Disclosed variant
已公開變題：把第二小節左手改為 F3 兩拍、C3 兩拍。這次音高與節奏都變；右手仍不變。先寫預測，此例不是保密遷移題。 Disclosed variant: bar 2’s left hand becomes F3 for two beats, then C3 for two. Both pitch and rhythm now change; the right hand does not. Predict first; this is not a hidden transfer test.
完成檢核 / Completion check
更正一個具體錯誤、數對事件、拒絕保證，並如實記錄自己是否真的合手試過。 Correct a concrete error, count events, reject guarantees, and truthfully record whether hands-together playing was tried.
3. Q+A | 音樂性和混音效果，誰比較重要？
Q+A: Musicality or mixing—which matters more?
WEB-0937803de260 · 原文 / Source
頁面日期 / Page date: 2025-01-21 · 標示篇幅 / Labeled length: 688 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 32–52；擷取總行 / extraction total: 64.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
依交付方式選記譜或音訊流程；先處理編寫與演奏問題，不把音樂性與混音對立。 Choose notation or audio workflows by delivery needs; address writing and performance early rather than opposing musicality to mixing.
用途 / Role
分辨問題階段。 Locate the problem stage.
來源限制與編輯保留 / Source limitations and editorial qualifications
作者流程偏好，非排他規則。 Author workflow preferences, not exclusive rules.
來源衍生預算 / Source-derived budget: 127/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: 無 / none.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
三張交付卡，找對要修改的東西 Three delivery cards: edit the right object
以同一個原創微型譜面，分清意圖、演奏紀錄與錄音效果主張，不用混音代替缺失的證據。 Use one original miniature score to distinguish intent, a performance record, and recording-effect claims without treating mixing as missing evidence.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 紙上題，不播放。4/4，C4＝中央 C；右手單音，左手休止，無踏板。事件清單是題目指定，不是音訊辨識結果。 Paper only; no playback. In 4/4, C4 = middle C. Right-hand single notes, left hand resting, no pedal. Event lists are stipulated, not extracted from audio.
意圖事件 / Intended events1.小節 / Bar 1
拍位 / Beat 1
音高 / Pitch E4
時值（拍）/ Duration (beats) 1
手部 / Hand 右手 / Right hand
2.小節 / Bar 1
拍位 / Beat 2
音高 / Pitch D4
時值（拍）/ Duration (beats) 1
手部 / Hand 右手 / Right hand
3.小節 / Bar 1
拍位 / Beat 3
音高 / Pitch C4
時值（拍）/ Duration (beats) 2
手部 / Hand 右手 / Right hand
卡片 / Cards1.標記 / ID P
描述 / Description 交付「給人彈的譜」；意圖結尾是 C4 兩拍，但草稿最後寫 B3 兩拍。 Deliverable: a score for a player. The intended ending is C4 for two beats, but the draft ends with B3 for two beats.
2.標記 / ID Q
描述 / Description 譜正確；另有可靠人工動作紀錄指出最後按了 B3。沒有錄音，也沒說按錯的原因。 The score is correct; a separate reliable human action log states that the last key was B3. There is no recording or explanation of the cause.
3.標記 / ID R
描述 / Description 譜與動作紀錄均符合意圖；有人說一個未知錄音版本「太遠」。未提供音訊、房間、裝置或處理史。 Score and action log match the intent. Someone calls an unknown recorded version “too distant.” No audio, room, device, or processing history is supplied.
回答欄 / Response columns 目前已知｜首先核對的對象｜仍未知｜禁止的結論 Known now | first object to check | unknown | prohibited conclusion
先預測與作答 / Predict and try first
先對三張卡作答，再讀答案。只根據給定資料，不臆測手型、麥克風或效果設定。
Respond to all three cards before reading the key. Use only supplied facts; do not invent hand posture, microphone details, or effect settings.
對 P 畫出唯一需改的音符；對 Q 寫出「目標鍵／已知按鍵」；對 R 寫一個資料需求，不指定購物。
For P, mark the sole note to edit; for Q, write “target key / known pressed key”; for R, request evidence without prescribing a purchase.
判斷這句話是否有根據：「R 的錄音聽起來遠，因此把譜末的 C4 改高八度就修好了。」
Judge this claim: “R’s recording sounds distant, so moving the written final C4 up an octave fixes it.”
已公開答案與推理 / Disclosed answers and reasoning
P：把最後 B3 改成 C4，時值不變；改譜即可滿足題給的書面意圖，但尚未聽到任何演奏。
P: replace final B3 with C4 without changing duration. This meets the stated written intent, but no performance has been heard.
Q：目標 C4，動作紀錄 B3；可先核對該位置的讀譜與鍵位，不由一個事件診斷動作原因或琴技程度。
Q: target C4, logged key B3. First check score reading and key location at that point; one event does not diagnose movement cause or skill level.
R：先要求版本、原件與處理史，保留主觀描述。現有資料無法確定問題來自演奏、房間、錄音或效果，也不支持改八度。
R: request version identity, original, and processing history while retaining the subjective description. The data cannot locate the cause in performance, room, recording, or effects, and do not support an octave change.
任務界線 / Task limits
不主張任何軟體只能作某一步；「先核對」不等於不同工作必須線性完成。「遠」沒有在此被量測。 No software is restricted to a single stage; a first check does not require a strictly linear workflow. “Distant” is not measured here.
已公開變題 / Disclosed variant
已公開變題：若可靠更正指出 Q 的動作其實是 C4，就撤回「按錯鍵」判斷；不要為保住原答案而假設新的錯誤。 Disclosed variant: if a reliable correction says Q’s actual key was C4, withdraw the wrong-key judgment instead of inventing a new error to preserve it.
完成檢核 / Completion check
P 改對符號，Q 不過度診斷，R 能停在缺證；不把聲音形容詞當配器或琴技的直接證據。 Correct P’s symbol, avoid overdiagnosing Q, and preserve R’s missing evidence; do not treat sound adjectives as direct proof of instrumentation or playing skill.
4. 只用「一台鍵盤」要怎麼彈出豐富聲響？
Rich sounds from one keyboard
WEB-12ba57b6ebf3 · 原文 / Source
頁面日期 / Page date: 2023-05-20 · 標示篇幅 / Labeled length: 1979 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 30–136；擷取總行 / extraction total: 150.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
介紹重疊、分割、力度切換、音程疊加、琶音器與打擊觸發。 Introduces layering, splits, velocity switching, interval layers, arpeggiation, and percussion triggering.
用途 / Role
區分鍵與聲部。 Distinguish keys and voices.
來源限制與編輯保留 / Source limitations and editorial qualifications
非原聲琴功能；即時演奏僅據文字陳述。 Not acoustic-piano features; live performance is only reported in text.
來源衍生預算 / Source-derived budget: 142/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: original_feedback_app_acceptance_design.acceptance_cases[0].given.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
四次按鍵，幾個輸出標籤？ Four key events: how many output labels?
用虛構路由表，辨認按鍵、音高和聲部標籤不是同一個計數，並避免把電子設定強加給原聲鋼琴。 Use a fictional routing table to distinguish key, pitch, and voice-label counts without imposing electronic settings on an acoustic piano.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 純紙上模型，不使用電子琴、軟體或錄音。C4＝中央 C；四個輸入事件各一拍，依次開始且不重疊。每個事件明定只有一次按鍵。 Paper model only, with no electronic keyboard, software, or recording. C4 = middle C. Four input events each last one beat, start sequentially, and do not overlap. Each explicitly contains exactly one key press.
輸入事件 / Input events1.拍位 / Beat 1
琴鍵 / Key D4
2.拍位 / Beat 2
琴鍵 / Key F4
3.拍位 / Beat 3
琴鍵 / Key A4
4.拍位 / Beat 4
琴鍵 / Key F4
虛構路由 / Invented router 題定虛構規則：低於 G4 的輸入只產生標籤 P，音高不變；G4 或以上同時產生 P 與 S，兩者音高都等於輸入。標籤不是已聽到或已核實的樂器。 Stipulated fictional rule: below G4, produce only label P at the input pitch; at G4 or above, produce P and S simultaneously, both at the input pitch. Labels are not heard or verified instruments.
計數規則 / Counting rule 分別數：輸入按鍵事件、輸出標籤事件、不同的「拍位＋音高」組合。兩標籤若同拍同音，第三欄只算一個組合。這不是聲學泛音模型。 Count separately: input key events, output-label events, and distinct beat-plus-pitch pairs. Two labels at the same beat and pitch count as one pair in the third column. This is not an acoustic-partials model.
先預測與作答 / Predict and try first
先逐拍展開路由，不看答案：哪一拍有兩個標籤？預測三欄總數。
Before the key, expand routing beat by beat: which beat has two labels? Predict all three totals.
判斷「五個輸出標籤，等於演奏者按了五次鍵」是否成立；引用本題輸入表而不是想像演奏影片。
Decide whether “five output labels mean five key presses” follows; cite the given input table instead of imagining a performance video.
可選原聲琴對照：左手休止，右手 D4、F4、A4、F4 每音一拍，無踏板。這只是四次實際按鍵，不聲稱觸發 P/S 或模仿額外樂器。
Optional acoustic comparison: left hand rests while the right plays D4, F4, A4, F4, one beat each, with no pedal. These are four actual presses, not claimed P/S triggering or extra-instrument imitation.
已公開答案與推理 / Disclosed answers and reasoning
第一拍 P:D4；第二拍 P:F4；第三拍 P:A4 與 S:A4；第四拍 P:F4。輸入＝4，標籤事件＝5，拍位＋音高組合＝4。
Beat 1 P:D4; beat 2 P:F4; beat 3 P:A4 and S:A4; beat 4 P:F4. Inputs = 4; label events = 5; beat-plus-pitch pairs = 4.
五個標籤不推出五次按鍵；但這也不證明任何真實錄音用了一個人或一台琴。真實情況需要另有按鍵或工程證據。
Five labels do not imply five key presses. Nor does this establish that any real recording used one person or one keyboard; that would need separate action or project evidence.
任務界線 / Task limits
路由是本題假設，不是對產品或真實演奏的測試。音高相同不代表音色、相位、音量或聽感相同；這些都未建模。 Routing is stipulated, not a product or performance test. Equal pitch does not imply equal timbre, phase, level, or perception; those properties are unmodeled.
已公開變題 / Disclosed variant
已公開變題：第三拍改 F4；四拍都只產生 P，三欄總數均為 4。先預測再看此答案；不是未知測試集。 Disclosed variant: change beat 3 to F4. Every beat now produces only P, so all three totals are 4. Predict before this answer; it is not an unseen test set.
完成檢核 / Completion check
能給出 4／5／4 及理由，也能說明自己的原聲琴不需任何對應設定。 Give and explain 4/5/4, and state that the acoustic piano needs no corresponding setting.
5. 4 分鐘內換 10 種音樂風格！我的配樂設計過程大公開～
Ten styles in four minutes: a scoring-design walkthrough
WEB-b8680c8fe2e7 · 原文 / Source
頁面日期 / Page date: 2022-04-09 · 標示篇幅 / Labeled length: 2432 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 32–120；擷取總行 / extraction total: 128.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
個案以重用旋律連接不同配器，手工對準指定畫面事件，結尾重現前段素材。 This case reuses melodies across orchestration changes, manually aligns selected visual events, and recalls earlier material at the ending.
用途 / Role
規劃連續與變化。 Plan continuity and change.
來源限制與編輯保留 / Source limitations and editorial qualifications
風格聯想屬作者選擇；原始工程未查。 Style associations are the author’s choices; project files uninspected.
來源衍生預算 / Source-derived budget: 142/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: 無 / none.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
三格小故事：不變的線與變化的底 Three-panel story: a fixed line over a changing base
保持同一原創旋律，明確規劃另一聲部的變化與一個畫面定位點；不靠風格或民族標籤猜聽感。 Keep an original melody fixed while planning another part’s change and one visual cue, without predicting perception from genre or cultural labels.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 紙上先做，可選原聲琴。三小節 4/4；C4＝中央 C；無踏板。右手每小節 G4 一拍、E4 一拍、D4 兩拍。右手聲部不省略；左手如下，兩手按拍位同時進行。 Start on paper; acoustic piano is optional. Three bars of 4/4; C4 = middle C; no pedal. In every bar, the right hand plays G4 for one beat, E4 for one, D4 for two. This melody is not omitted; the left hand below runs simultaneously by beat.
小節表 / Bar plan1.小節 / Bar 1
標籤 / Label A：保持 A: hold
左手事件 / Left-hand events1.拍位 / Beat 1
音高 / Pitch C3
時值（拍）/ Duration (beats) 4
2.小節 / Bar 2
標籤 / Label B：重新起音 B: reattack
左手事件 / Left-hand events1.拍位 / Beat 1
音高 / Pitch C3
時值（拍）/ Duration (beats) 1
2.拍位 / Beat 2
音高 / Pitch C3
時值（拍）/ Duration (beats) 1
3.拍位 / Beat 3
音高 / Pitch C3
時值（拍）/ Duration (beats) 1
4.拍位 / Beat 4
音高 / Pitch C3
時值（拍）/ Duration (beats) 1
3.小節 / Bar 3
標籤 / Label A′：回到保持 A′: return to hold
左手事件 / Left-hand events1.拍位 / Beat 1
音高 / Pitch C3
時值（拍）/ Duration (beats) 4
紙上畫面 / Storyboard 三張自畫紙卡只寫「門關著」「門打開」「門關上」。唯一同步目標：第三張切換在第三小節第一拍。這是虛構提示，不是作者影片內容。 Three self-drawn paper cards read only “door closed,” “door opens,” and “door closes.” Sole synchronization target: switch to card 3 at bar 3 beat 1. These are invented cues, not the source video.
紙上時間規則 / Paper time rule 若另假設每拍恰為一秒、第一拍起點 t＝0，才可換算紙上秒數；實際試彈不要求計時，也不需節拍器。 Only if one additionally stipulates exactly one second per beat and the first onset at t = 0 may paper seconds be calculated. Playing does not require timing or a metronome.
先預測與作答 / Predict and try first
先預測：右手哪一格有改？第二小節左手起音數增加多少？第三張的目標是全段第幾拍？
Predict first: which right-hand cell changes? How many left-hand attacks are added in bar 2? Which global beat is card 3’s target?
寫兩欄「保持／改變」，再依給定的一秒一拍假設算出第三小節開始的 t；不要把算術當成實際同步已成功。
Write “fixed / changed” columns, then calculate bar 3’s start time under the stipulated one-second beat. Do not treat arithmetic as proof of successful real synchronization.
可選：先分手再合手，慢慢試三小節；需要時小節間停下，並記「含停頓」，不要仍聲稱演奏十二秒。
Optional: try hands separately, then together, slowly. Pause between bars if needed and label the attempt “with pauses”; do not still claim a twelve-second performance.
寫自己的聽感，允許「只是重複變多，沒覺得故事更清楚」。不用把 B 命名為特定流派，也不用猜樂器編制。
Write your own impression, allowing “more repetitions, but no clearer story.” B needs no genre label or inferred instrumentation.
已公開答案與推理 / Disclosed answers and reasoning
右手完全不變，各小節三次起音，共九次。左手第一／二／三小節是 1／4／1 次；第二小節比第一小節多三次。
The right hand is unchanged: three attacks per bar, nine total. Left-hand attacks are 1/4/1 by bar; bar 2 has three more than bar 1.
第三小節第一拍是全段第九拍；在指定時間模型中 t＝8 秒，全段三小節長 12 秒。沒有計時的演奏不能驗證這些秒數。
Bar 3 beat 1 is global beat 9. In the stipulated time model, t = 8 seconds and the three bars last 12 seconds. Untimed playing cannot validate those seconds.
已知的是音符與事件表。旋律重複是否帶來連貫感、各人是否聽出畫面，都還沒被證明。
What is known is the note/event plan. Whether repetition produces coherence or listeners infer the pictures remains unproven.
任務界線 / Task limits
三個織度標籤不是三個完整音樂風格。C3 配 D4 的持續同響只是本題指定，並不保證所有人覺得舒適或適合結尾。 Three texture labels are not three full musical styles. Sustained C3 with D4 is stipulated, not guaranteed comfortable or suitable as an ending for everyone.
已公開變題 / Disclosed variant
已公開變題：若第二小節刪去左手第三、第四拍，第二小節變兩次起音，第四拍仍有右手 D4 持續；不可稱最後兩拍全體休止。 Disclosed variant: remove left-hand beats 3 and 4 in bar 2. That bar has two left-hand attacks, but right-hand D4 still sustains through beat 4; do not call the last two beats total silence.
完成檢核 / Completion check
能分清兩手、指出第九拍與 t＝8 秒的假設，並保留聽感與實際演奏的未知。 Distinguish the hands, identify beat 9 and the assumption behind t = 8 seconds, and retain uncertainty about perception and actual performance.
6. 來做一首有太空船感覺的歌！然後給你七個好用的編曲秘訣～
A spaceship-like song: seven arranging ideas
WEB-2644a9209109 · 原文 / Source
頁面日期 / Page date: 2022-06-06 · 標示篇幅 / Labeled length: 3447 中文字 / Chinese characters.
已讀正文行 / Reviewed body lines: 30–147；擷取總行 / extraction total: 155.
完整正文；媒體未審。 Complete body; media unreviewed.
作者說明。 Author exposition.
來源重點 / Source insight
以較高和弦度數重配旋律，合音看當下和聲；個案也談音效、鼓、處理、空間與手工節奏修整。 Reharmonizes melody as higher chord degrees and considers the current harmony for backing voices; also discusses effects, drums, processing, ambience, and manual timing edits.
用途 / Role
分開音符與意象。 Separate notes from imagery.
來源限制與編輯保留 / Source limitations and editorial qualifications
太空感與節奏微調偏好非通則；產品現況未核實。 Space imagery and timing preferences are not universal; current products unverified.
來源衍生預算 / Source-derived budget: 159/200（兩語合計 / both languages combined）。
計入欄位 / Counted fields: title_zh, title_en_editorial, visible_date, visible_length_zh_characters, coverage.reviewed_text_lines, coverage.extraction_total_lines, coverage.note, source_claim_summary, content_role, evidence_type, limits.
跨段落計入 / Cross-section counted paths: 無 / none.
任務的音符、數據、假想主張、步驟與答案均為新設計；不是來源練習或歌曲的改寫。 Task notes, data, hypothetical claims, steps, and answers are newly designed, not adaptations of a source exercise or song.
原創任務 / Original task
同一個上方音，兩張和聲卡 One upper note, two harmony cards
在完整給定音高的微型例子裡核對音名關係，再把「太空感」留作個人描述，而不是音符的必然性質。 Check pitch-name relations in fully specified miniature examples, then leave “space-like” as a personal description rather than an inevitable property of notes.
原創教材草案；沒有實際演奏、使用者測試或學習成效驗證。 Original teaching draft; no actual performance, user testing, or learning-outcome validation.
資料卡 / Data card
設定 / Setup 先做紙上集合比對。C4＝中央 C。A/B 各是一小節 4/4，四拍內所有列出音同時開始並持續，版本間停下。沒有任何未列旋律、低音、鼓或聲效。 First compare sets on paper. C4 = middle C. A/B are each one bar of 4/4; all listed notes start together and last four beats. Stop between versions. There is no unlisted melody, bass, drum, or effect.
卡片 / Cards1.標記 / ID A
左手音高 / Left-hand pitches["C3", "E3", "G3"]
右手音高 / Right-hand pitch E4
時值（拍）/ Duration (beats) 4
三和弦音名 / Triad pitch names["C", "E", "G"]
2.標記 / ID B
左手音高 / Left-hand pitches["D3", "F3", "A3"]
右手音高 / Right-hand pitch E4
時值（拍）/ Duration (beats) 4
三和弦音名 / Triad pitch names["D", "F", "A"]
給定名稱 / Given labels 本題稱 A 的左手集合 C 大三和弦，B 的左手集合 D 小三和弦；「集合內」只問音名，暫不管八度。比較這兩張完整卡，不還原任何來源歌曲。 The task names A’s left-hand set C major and B’s D minor. “In the set” asks about pitch names, temporarily ignoring octave. Compare these complete cards; reconstruct no source song.
可選試彈界線 / Optional playing boundary 若想試原聲琴，先確認左手三音可舒服同按，右手彈 E4，無踏板。不舒服就只做紙上題；若改成逐音彈，須標「琶音版」，不能當成原定同時版本的聽感證據。 If trying acoustic piano, first check that the left-hand triad is comfortable together; play E4 in the right hand with no pedal. Otherwise stay on paper. If notes are played one by one, label it “arpeggiated,” not evidence for the stipulated simultaneous version.
先預測與作答 / Predict and try first
在讀答案前，預測兩版上方音是否相同；分別圈出 E 是否出現在左手的音名集合。
Before reading the key, predict whether the upper note is unchanged; circle whether E appears in each left-hand pitch-name set.
逐音核對兩張卡；寫出固定項與改變項。不要從「有四個音」就推論四件樂器。
Check each pitch and list fixed versus changed items. Do not infer four instruments merely from four notes.
若選擇試彈，分別記錄你如何演奏、自己的描述或「未聽出差別」。紙上答案不依賴喜歡哪版，也不要求使用太空、魔幻等形容。
If playing, record how you performed it and your own description or “no difference noticed.” The paper answer depends on neither preference nor space/magic adjectives.
最後判斷：「只知道上方 E4，就能決定伴奏一定是 B」是否有根據。
Finally judge: “Knowing only upper E4 determines that the accompaniment must be B.”
已公開答案與推理 / Disclosed answers and reasoning
A/B 的上方音都是 E4。E 在 A 的 {C,E,G} 內，不在 B 的 {D,F,A} 內；這是音名集合關係，不是好聽分數。
Both upper notes are E4. E belongs to A’s {C,E,G}, not B’s {D,F,A}. This is pitch-name membership, not a beauty score.
固定：上方音、起拍、長度及每版列出四個音；改變：三個左手音。只有音符表，沒有樂器錄音證據。
Fixed: upper note, onset, duration, and four listed notes per version. Changed: all three left-hand pitches. A note table supplies no recorded-instrument evidence.
單獨 E4 不足以唯一決定伴奏；兩張卡本身已給出不同的可列寫例子。不能從音名集合推出人人的意象或偏好。
E4 alone does not uniquely determine accompaniment; the two cards themselves supply different writable examples. Pitch-name sets do not establish universal imagery or preference.
任務界線 / Task limits
這不是完整重配和聲分析或來源譜例轉錄，也未測同步演奏效果。沒有理由為了特定意象購買音色或把節奏彈不準。 This is neither complete reharmonization analysis nor a source-score transcription, and simultaneous performance is untested. It gives no reason to buy sounds or play inaccurately to achieve imagery.
已公開變題 / Disclosed variant
已公開變題：右手都改 F4。F 不在 A 左手集合，但在 B 集合；仍不能推出哪版更好。兩個版本均為已公開示範，不是未見遷移測驗。 Disclosed variant: change both right hands to F4. F is absent from A’s left-hand set but present in B’s; superiority still does not follow. Both versions are disclosed demonstrations, not unseen transfer tests.
完成檢核 / Completion check
能正確圈集合、保留缺失資訊，且不把和聲標記、形容詞或音符數當成聽感保證或配器證明。 Mark membership correctly, retain missing information, and avoid treating harmony labels, adjectives, or note counts as perception guarantees or instrumentation proof.
未來回饋工具：原創驗收草案 / Future feedback tool: original acceptance draft
原創需求與驗收草案；未建置或測試應用程式，不宣稱模型通過。 Original requirement/acceptance draft; no application was built or tested and no model is claimed to pass.
必需欄位 / Required fields
證據範圍：文字／音訊／影片／圖片／工程分開；保存原網址、日期、可見部分與未讀部分。
Evidence scope: separate text/audio/video/image/project review; retain original URL, date, visible coverage, and unread material.
題目真值：音高、拍位、時值、手部、是否同時；演奏版本與紙上版本不要混為一談。
Task ground truth: pitches, beats, durations, hands, and simultaneity; keep paper and played versions distinct.
預測與觀察分欄；答案曝光與變題曝光明記；偏好不填成正確率或能力分數。
Separate predictions from observations; record answer/variant exposure; do not encode preference as accuracy or ability.
驗收案例 / Acceptance cases
C16-01
來源只有文字寫著一人一琴同時演奏。 Only a textual claim of one-person, one-keyboard live performance.
顯示「作者陳述，未看示範」，不標影音已審。 Show “author-reported; demonstration unreviewed,” not media-reviewed.
C16-02
某份 2025 年文章提出 AI 工具評價。 A 2025 article evaluates AI tools.
保留日期與歸屬，不轉成今日能力排行。 Retain date and attribution, not a current capability ranking.
C16-03
四次輸入對應五個虛構輸出標籤。 Four inputs map to five fictional output labels.
保留不同計數，不判多彈一次或五件樂器。 Preserve distinct counts; infer neither an extra press nor five instruments.
C16-04
使用者只分手試過三小節，且有停頓。 The user tried three bars only hands separately and with pauses.
不宣稱完整合手、十二秒同步或學習遷移。 Claim neither full hands-together performance, twelve-second synchronization, nor learning transfer.
C16-05
兩個配法同用 E4，但只有音名表。 Two settings share E4, with only pitch-name tables supplied.
可比較集合；不可推太空感、配器或哪個更好。 Compare sets; do not infer space imagery, instrumentation, or superiority.
若系統把來源日期抹掉、暗示看過未讀媒體、捏造使用者演奏、將任務答案當未見測試，或要求為原聲琴購買電子功能，則不接受。 Reject if a system removes source dates, implies unseen-media review, invents user playing, treats disclosed answers as unseen tests, or requires electronic purchases for the acoustic piano.
初學順序 / Beginner route
先做第 1、3 題：只查修改座標與證據類型。
Start with tasks 1 and 3: edit coordinates and evidence types only.
再做第 2、4 題：核對建議、音符與路由計數。
Then tasks 2 and 4: advice, note, and routing checks.
最後做第 5、6 題：兩手時間線與和聲卡；紙上完成即可，試彈非先決條件。
Finish with tasks 5 and 6: hand timelines and harmony cards; paper completion is sufficient, and playing is not a prerequisite.
核對狀態 / Validation status
本機 Python 只驗證欄位、符號音高、時值、事件數、集合與預算；沒有執行模型推論、播放、琴上測試或成效試驗。 Local Python checked only fields, symbolic pitches, durations, event counts, sets, and budgets; no model inference, playback, piano testing, or efficacy experiment.
六組 ID／網址／原題與指派完全一致；兩語欄位完整；未審媒體旗標為否。
All six ID/URL/title triples match assignment; bilingual fields are complete; unreviewed-media flags are false.
六篇來源衍生文字含指定元資料與跨段落重述，合計為 119、183、127、142、142、159 單位，均不超過硬性上限 200。
Source-derived text, including specified metadata and cross-section restatements, totals 119, 183, 127, 142, 142, 159 units; all meet the hard 200 cap.
第 1 題八拍、七音事件，末拍休止變六；第 2 題左手 2→5 次；第 3 題總時值四拍。
Task 1 has eight beats/seven note events, falling to six with the final rest; task 2 left-hand counts change 2→5; task 3 durations total four beats.
第 4 題路由算得 4／5／4；第 5 題左手 1／4／1、右手共九次，提示點第九拍＝假設時間 8 秒。
Task 4 routing yields 4/5/4; task 5 left-hand counts are 1/4/1 and right-hand total nine, with cue beat 9 = hypothetical time 8 seconds.
第 6 題 E 集合關係真／假，變題 F 為假／真；各版本四音、四拍，左手範圍均七半音。
Task 6 E membership is true/false, with variant F false/true; each version has four notes/four beats and a seven-semitone left-hand span.
六篇均無存取阻礙；本成果沒有進行播放、原聲鋼琴演奏或應用程式驗證。 All six were accessible; this deliverable includes no playback, acoustic-piano performance, or application validation.