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.
第十八批：軟體與工具觀念 / Batch 18: Software and tool concepts
六篇公開工具文章的歷史閱讀卡；研究資料表述與證據邊界，不建置工作站。
Historical reading cards for six public tool articles; research representations and evidence boundaries, not workstation setup.
閱讀範圍與限制 / Scope and limits
- 完整保留來源 ID、網址與中文原題；英文題名為編輯翻譯。兩個檔案是同一成果，不重複計數。
Preserve source IDs, URLs, and original Chinese titles; English titles are editorial translations. Both files represent one deliverable, counted once.
- 來源摘要均描述發表當時作者的說法。產品版本、平台、價格、指令、模型能力、音質與效率不視為目前已驗證的事實或建議。
Source summaries describe the author’s publication-time account. Product versions, platforms, prices, commands, model capabilities, sound quality, and efficiency are not treated as currently verified facts or recommendations.
- 本批只讀公開文字；沒有聆聽、播放、執行來源程式或音樂軟體、安裝、購買、下載媒體、上傳、推送或啟動本機模型。連結課程、圖表與內嵌媒體未審。
Only public text was read. No listening, playback, source-program or music-software execution, installation, purchase, media download, upload, push, or local-model session occurred. Linked courses, figures, and embedded media remain unreviewed.
- 不抄錄來源旋律、歌詞、樂譜、示範程式或安裝指令，也不推論未聽示範的聲音。本文不提供規避存取限制的方法。
No source melody, lyrics, score, demonstration code, or installation command is copied, and unheard examples are not reconstructed. No access-restriction bypass instructions are provided.
- 紙上資料練習是獨立原創設計，不是任何軟體語法或產品功能承諾。它不要求唱歌、麥克風、帳號、電子琴或新增器材。
Paper-data exercises are independent original designs, not any tool’s syntax or a product capability promise. They require no singing, microphone, account, digital piano, or new equipment.
- 來源衍生文字按每個中文字元及英數詞元合併計數；含雙語題名、摘要、用途、限制、涵蓋與重述，嚴格不超過 200。
Source-derived text is counted by CJK characters plus alphanumeric tokens across both languages, including titles, summaries, roles, limits, coverage, and restatements, strictly capped at 200.
- 讀懂工具文章不能證明學會鋼琴；本批沒有學習者測試、模型基準或音訊準確率，不能聲稱效果成立。
Understanding a tool article does not establish piano learning. This batch contains no learner study, model benchmark, or audio-accuracy measurement and establishes no efficacy claim.
- 未審核任何外部素材的使用許可。後續若要處理真實檔案，應先另行確認來源、用途與授權範圍；本批不給法律意見。
No external asset’s usage permission was audited. Any future real-file processing first needs separate confirmation of provenance, purpose, and authorization scope; this batch provides no legal advice.
1. Aiko 第二代好好用！怎麼讓她唱得更像真人？以及好和弦居然用 24 年前的老機器在作曲？ / Aiko’s second generation: singing more like a human, and composing with a 24-year-old device
來源 / Source: WEB-88fabe2855f9 · 原文 / Primary article
日期 / Date: 2021-05-07 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 1878
涵蓋行 / Covered lines: 30–100 / 102 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
作者調整音符銜接、音高曲線與音素時長，並使用舊音源器製作伴奏。
The author edits note connections, pitch curves, and phoneme durations, using an older sound module for accompaniment.
用途 / Role
表情參數分層。
Separate expression parameters.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
擬人聽感未驗證；不把歌聲滑音當鋼琴技巧。
Human likeness unverified; vocal slides are not piano technique.
雙語來源衍生計數 / Combined source-derived units: 166 / 200
2. 超～極簡系音樂製作法！只要打打字就可以做出音樂？ / Ultra-minimal music making: composing by typing
來源 / Source: WEB-c4243a139828 · 原文 / Primary article
日期 / Date: 2021-08-20 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 7393
涵蓋行 / Covered lines: 30–618 / 621 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
Alda 用文字指定音高、時值、聲部、反覆與移調；括組可按比例縮放節奏。
Alda text specifies pitch, duration, voices, repetition, and transposition; grouped rhythms can be proportionally rescaled.
用途 / Role
可檢查的事件表述。
Inspectable event representation.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
語法與安裝屬當年；未執行範例或複製曲譜。
Syntax and setup are historical; examples unexecuted and scores uncopied.
雙語來源衍生計數 / Combined source-derived units: 127 / 200
3. 來用 Linux 和自由軟體做音樂吧！ / Making music with Linux and free software
來源 / Source: WEB-3aea319f315f · 原文 / Primary article
日期 / Date: 2021-08-12 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 3913
涵蓋行 / Covered lines: 30–162 / 164 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
以舊電腦創作為題：Ardour 錄混音、Qtractor 編序、Helm 合成、MuseScore 製譜，也列商業選項。
For older-computer creation: Ardour recording/mixing, Qtractor sequencing, Helm synthesis, and MuseScore notation; commercial options also appear.
用途 / Role
按工作分工具。
Classify tools by task.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
權利說法須補正，見 SUP18-FREE；品質未驗證。
Rights wording needs correction: SUP18-FREE; quality unverified.
雙語來源衍生計數 / Combined source-derived units: 115 / 200
4. MilkyTracker - 你一定沒聽過但超大推的編曲軟體 | 好軟 NiceSoft / MilkyTracker: an overlooked arranging tool | NiceSoft
來源 / Source: WEB-b9e215217ae2 · 原文 / Primary article
日期 / Date: 2022-07-06 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 3979
涵蓋行 / Covered lines: 30–188 / 191 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
直向表格編排取樣；音符、取樣與音高對應分開，片段由播放清單串接。
Vertical tables sequence samples; notes, samples, and pitch mapping are distinct, with patterns assembled through an order list.
用途 / Role
區分事件與音源。
Separate events from sound sources.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
介面屬當年；可取得的取樣未必可任意再用。
Interface is historical; accessible samples are not automatically reusable without restriction.
雙語來源衍生計數 / Combined source-derived units: 130 / 200
5. 一次了解！在家裡製作音樂需要的基本配備～（2021 版） / Basic home music-production equipment: the 2021 edition
來源 / Source: WEB-3ac9f4061831 · 原文 / Primary article
日期 / Date: 2021-09-03 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 4842
涵蓋行 / Covered lines: 41–207 / 216 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
分辨控制器與可獨立發聲鍵盤；錄音介面需求看同時輸入，收音設備視錄製內容而定。
It distinguishes controllers from self-sounding keyboards; interface needs depend on simultaneous inputs and capture needs on the material.
用途 / Role
按任務辨需求。
Identify task-specific needs.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
採購與延遲判斷屬當年經驗，不是本案清單。
Purchasing and latency judgments are historical experience, not this project’s checklist.
雙語來源衍生計數 / Combined source-derived units: 140 / 200
6. 你的 D 槽好夥伴：youtube-dl、ffmpeg、sox……音樂人一定要會的工具！ / Media tools for musicians: youtube-dl, FFmpeg, and SoX
來源 / Source: WEB-d6a49faebae0 · 原文 / Primary article
日期 / Date: 2021-02-13 · 讀取 / Accessed: 2026-10-22 · 頁示字數 / Displayed length: 3141
涵蓋行 / Covered lines: 30–180 / 182 · reviewed_public_article_text_only
涵蓋 / Coverage
完整正文；媒體未審。
Complete body; media unreviewed.
來源要點 / Source ideas
以 youtube-dl 取得媒體、FFmpeg 轉檔、SoX 改速度與音高，區分三類工作。
It separates three tasks: youtube-dl media retrieval, FFmpeg conversion, and SoX speed/pitch changes.
用途 / Role
區分取得與加工。
Separate acquisition from transformation.
證據 / Evidence
作者教學。
Author exposition.
限制 / Limits
舊指令與支援未驗證；不提供下載或繞過限制步驟。
Old commands and support unverified; no download or bypass steps supplied.
雙語來源衍生計數 / Combined source-derived units: 129 / 200
補充核對 / Supplementary verification
什麼是自由軟體？ / What is Free Software? — GNU Project / Free Software Foundation
SUP18-FREE · 官方來源 / Official source · 2026-10-22
官方索引文字的定義與商業段落；正文開啟兩次逾時，未讀完整頁。
Official indexed text: definition and commercial sections. Two direct opens timed out; full page unread.
自由指使用者權利，不是零售價；自由軟體可商用，許多授權仍以版權為基礎。
Freedom concerns user liberties, not zero price; free software can be commercial and many licenses still rely on copyright.
不等於公有領域；非個案授權審核或法律意見。
Not equivalent to public domain; no individual-license audit or legal advice.
雙語來源衍生計數 / Combined source-derived units: 139 / 200
原創研究整合 / Original research synthesis
原創資料模型：分開「預定音樂事件」「產生的示範」「觀察到的演奏」「觀察信心」。缺少觀察時，顯示未測，不由示範好聽推論學生會彈。
Original data model: separate intended musical events, generated demonstrations, observed performance, and observation confidence. When observation is absent, show unmeasured rather than infer player skill from a pleasing demonstration.
同一練習保留單一中立資料表，再由它產生中英說明；兩語共享事件 ID、音高、拍位、時值與答案版本。不能由翻譯另造一套數字。
Keep one neutral table per exercise and derive Chinese and English instructions from it. Both languages share event IDs, pitches, beat positions, durations, and answer versions; translation must not create a second numerical truth.
人工修訂要保留修訂前值、修訂後值、原因與操作者。任何推測只作候選，不覆蓋學習者輸入；未確定可以是正式結果。
Human corrections retain before/after values, reason, and editor. Inferences remain candidates rather than overwriting learner input; unresolved is an acceptable final result.
概念卡與鋼琴練習分開：工具操作沒有自然對應的手部技能，就保留閱讀與判斷題，不硬做成技巧課。
Separate concept cards from piano practice: when tool operation has no natural motor-skill counterpart, retain reading and reasoning rather than forcing a technique lesson.
原創紙上題 / Original paper exercise: 事件表不是演奏證明 / An event table is not proof of performance
用紙筆驗證資料一致性及證據分類；不測軟體、不評演奏好壞。
Use paper to check data consistency and evidence classification, not software behavior or playing quality.
紙筆與以下四列。拍位從 0 算；1 拍為四分音符，4 拍是一個 4/4 小節；每列依序開始。
Paper and the four rows below. Positions begin at 0; one beat is a quarter note and four beats form one 4/4 bar; rows start in sequence.
| 事件識別 / Event identifier | 音高；C4 為中央 C / Pitch; C4 is middle C | 開始拍位 / Start position in beats | 持續拍數 / Duration in beats |
|---|---|---|---|
| P1 | C4 | 0 | 1 |
| P2 | E4 | 1 | 1 |
| P3 | D4 | 2 | 1 |
| P4 | C4 | 3 | 1 |
步驟 / Steps
- 先遮住答案。計算每列結束拍位；判斷表中相鄰事件是否有空隙或重疊。
Cover the answer first. Calculate each end position and determine whether adjacent table events have gaps or overlaps.
- 假設英文版把 P3 寫成開始拍位 2.5，其他資料不變。先寫出衝突，再選擇修正文字還是修改共同資料表；不要默默接受兩套版本。
Suppose the English version gives P3 a start of 2.5 while everything else stays unchanged. Identify the conflict, then choose whether the prose or shared table needs correction; do not silently accept two versions.
- 把三張紀錄分成預定、產生、觀察：A 是上表；B 說「系統產生了一份練習示範」；C 說「學生自述已完成」。哪些能支持「學生每音準確」？
Classify three records as intended, generated, or observed: A is the table; B says a system generated a practice demonstration; C says a learner reports completion. Which supports “the learner played every note accurately”?
參考答案 / Answer key
- 結束拍位為 1、2、3、4。依表定義，沒有空隙或重疊；這只是計畫數值，未描述實際聲音。
End positions are 1, 2, 3, and 4. By the table’s definitions there are no gaps or overlaps; these are planned numbers, not observed sound.
- 若共同表為權威且未批准改題，英文 P3 應改回 2。若採 2.5，P2 與 P3 間空 0.5 拍，P3 與 P4 重疊 0.5 拍；須先確認意圖。
If the shared table is authoritative and no exercise change was approved, restore English P3 to 2. At 2.5, P2–P3 have a 0.5-beat gap and P3–P4 a 0.5-beat overlap; first confirm intent.
- A 是預定；B 是產生；C 是自述，不是已核實的逐音觀察。三者皆不足以證明每音準確，應顯示「演奏未測」。
A is intended, B generated, and C self-report rather than verified note-level observation. None establishes every-note accuracy; display “performance unmeasured.”
可選原聲試彈 / Optional acoustic playing
可完全不彈。若已有原聲鋼琴且熟悉鍵位，可右手依序彈 C4、E4、D4、C4，每音一拍、共一小節；左手休息，不踩踏板，自選舒服速度。C4 是中央 C，往右白鍵依序 D4、E4。不需唱或錄音；不適即停止。這不模仿歌聲滑音。
Playing is optional. With an existing acoustic piano and familiar key locations, play C4, E4, D4, C4 in the right hand, one beat each for one bar; left hand rests, no pedal, at a comfortable chosen pace. C4 is middle C, followed rightward by white keys D4 and E4. No singing or recording; stop if uncomfortable. This does not imitate vocal pitch slides.
停止條件 / Stopping condition
列出四個結束拍位、找到翻譯衝突、把未測結果標清楚即可結束。已看答案就標記已見；不能把一次核對稱為學習增益。
Stop after listing four end positions, identifying the translation conflict, and labeling the unmeasured result. Mark the answer as seen if already read; one check is not evidence of learning gain.
原創應用驗收構想 / Original app acceptance design
SW18-01
情境 / Scenario
同一事件表的中英版本。
Chinese and English versions of one event table.
通過條件 / Pass condition
事件、數字與答案完全一致；文字差異不改變題意。
Events, numbers, and answers match exactly; wording differences do not change the task.
所需證據 / Evidence needed
對照表與人工雙語審核；尚未建置。
Comparison table and human bilingual review; not implemented.
SW18-02
情境 / Scenario
只有計畫及示範，沒有演奏觀察。
A plan and demonstration exist without a performance observation.
通過條件 / Pass condition
不能產生學生音準、節奏、觸鍵或踏板分數。
No learner pitch, rhythm, touch, or pedal score is generated.
所需證據 / Evidence needed
以虛構資料驗證規則，不能當音訊模型評測。
Verify the rule with fictional data, not as an audio-model evaluation.
SW18-03
情境 / Scenario
歷史閱讀卡被要求推薦現在的工具。
A historical reading card is asked to recommend a current tool.
通過條件 / Pass condition
標示需另作目前查證；不自動複製購買或安裝步驟。
Mark current verification as needed; do not automatically copy purchase or installation steps.
所需證據 / Evidence needed
日期與來源覆蓋稽核；本批不展開採購。
Date and source-coverage audit; purchasing is outside this batch.
SW18-04
情境 / Scenario
來源只有文字，但要求描述示範聽感。
Only text was reviewed, but an audible demonstration description is requested.
通過條件 / Pass condition
回答媒體未審，保留不確定，不補寫聲音。
State that media is unreviewed, preserve uncertainty, and do not invent sound.
所需證據 / Evidence needed
回覆與媒體狀態對照；不是盲聽實驗。
Compare the response with media-review status; not a blinded listening experiment.
跨批計數與查核 / Cross-batch counting and checks
已搜尋既有批次 JSON／Markdown 的六個精確主來源網址及 GNU 定義網址與辨識字串，未見先前內容重用。本批僅一項補充文字。未來跨批重述仍須按來源網址合併計數；雙格式只算一份。
Existing batch JSON/Markdown was searched for the six exact primary URLs and GNU definition URL/identifying strings; no prior substantive reuse was found. This batch uses one supplementary text. Future cross-batch restatements must aggregate by source URL; the two formats count once.
本機 Python 只驗證自製中繼資料、計數與算術；未執行來源程式或音樂軟體。
Local Python checked only original metadata, counts, and arithmetic; no source program or music software was executed.
六篇完整正文；媒體與產品執行皆為 0。來源識別、四個結束拍位與所有 200 單位上限均通過本機檢查。
Six complete article bodies; media reviews and product executions are all zero. Source identity, four end positions, and every 200-unit limit passed local checks.
直接開啟兩次；同網址官方搜尋文字可讀。
Two direct opens; official indexed text at the same URL was readable.
開啟逾時，非存取拒絕或用量耗盡。僅使用已讀索引文字，不宣稱完整正文。
Direct opens timed out, not access denial or usage exhaustion. Only read indexed text is used; full-body coverage is not claimed.
六篇正文已讀；雙語、來源與數值自檢完成，待獨立審核。
Six article bodies read; bilingual, source, and numerical self-checks complete, pending independent review.