Chapter I 台灣華語 Audiobook — Voice Bake-off
Status: Scratch AI narration for voice / tempo / style review only. Not the retail master (retail = human cast).
2026-08-02 lock: Albert + friends chose #2 HsiaoChen. Full Chapter I evaluation track: |Ch.1 ZH HsiaoChen full eval
Full EN SKU A (all chapters): |EN SKU A multi-track review hub
Identity Snapshot
- Work: Chapter I, Toward A Community of Hope (zh-TW AMT v1.0)
- Narration: AI scratch · three zh-TW neural voices · scholarly-pastoral register
- Purpose: Echopedia review of voice, tempo, and style before full zh-TW render
- Retail: human narrator; this build is not final master
Source AMT: v1.0 frozen · Chapter I — Formosan in Formosa Yesterday (zh-TW translation) Book: Dr. Albert S. Lai, Toward A Community of Hope
Voice Bake-off
Three zh-TW neural voices were rendered from the same 135-character opening excerpt (Section I — The Land — Formosa). All samples are loudnormed to -16 LUFS / -1.7 dBTP.
Samples
YunJheNeural (Male, Friendly/Positive) — rate +0%, pitch -1Hz
- Download MP3 · ~5:08 · 26.3 wpm
HsiaoChenNeural (Female, Friendly/Positive) — rate +0%, pitch +0Hz
- Download MP3 · ~5:48 · 23.3 wpm
HsiaoYuNeural (Female, Friendly/Positive) — rate +0%, pitch +0Hz
- Download MP3 · ~7:00 · 19.3 wpm
Narration QC
| Dimension | Choice | Why |
|---|---|---|
| Voices | YunJhe / HsiaoChen / HsiaoYu | All three zh-TW neural voices available via edge-tts |
| Tempo | +0% (default) | Scholarly-pastoral register naturally slower; dense geographical exposition benefits from measured pace |
| Pitch | YunJhe: -1Hz / HsiaoChen & HsiaoYu: +0Hz | YunJhe (male) gets slight gravitas; female voices at neutral |
| Loudness | -16 LUFS / -1.7 dBTP | Streaming audiobook reference (−16 to −18 LUFS, TP ≤ −1.5) — verified post-loudnorm on all three files |
| Style | Continuous prose, scholarly-pastoral | Academic tone without news-anchor delivery; suitable for 1971 dissertation content |
| Register | Taiwan Mandarin (zh-TW) | Per GLOSSARY.md: Formosa/Formosan policy — 1971 academic context retains Formosa/Formosan; general context uses Formosa（台灣） |
Bake-off summary
| Candidate | Duration | WPM | Notes |
|---|---|---|---|
| YunJheNeural | ~5:08 | 26.3 | Male voice; most measured pace; gravitas suits scholarly tone; pitch −1Hz adds warmth; TP −1.75 dB |
| HsiaoChenNeural | ~5:48 | 23.3 | Female voice; slightly slower; warm and clear; neutral pitch; TP −1.76 dB |
| HsiaoYuNeural | ~7:00 | 19.3 | Female voice; slowest; most deliberate; may be too slow for full chapter; TP −1.74 dB |
WPM context
The EN Ch.1 Christopher scratch measured ~149 wpm. The zh-TW voices are significantly slower (19–26 wpm) due to the denser character-per-word ratio of Chinese and the naturally measured scholarly-pastoral register. This is acceptable for review — the full zh-TW render (SKU B) may use rate +5% to +10% to reach a more comfortable ~30–35 wpm equivalent while maintaining the scholarly tone.
What to listen for (review checklist)
- Names/places: Formosa（台灣）, Taipei, Tainan, Pescadores（澎湖）, Kamchatka（堪察加） — natural stress and pronunciation?
- Geography density (first 5 min): Does tempo hold without rushing lists of distances and measurements?
- Pastoral vs academic: Does the voice feel appropriately scholarly without being news-anchor-like?
- Formosa/Formosan policy: Are the historical 1971 terms (Formosa, Formosan) read naturally without over-emphasis?
- Compare EN: |EN Ch.1 review (Christopher) — does the zh-TW voice carry the same scholarly authority?
Opening text (AMT excerpt)
第一章 — 昔日的 Formosa 之中的 Formosan
第一節 — 土地 — Formosa
這個叫做 Formosa 的島嶼，中國人和日本人都稱之為台灣。它是連接從卡馬楚卡（Kamchatka）到馬來半島的一串島嶼的其中一環，並且遮蔽著東亞大陸海岸，免受開闢的太平洋的衝擊。Formosa 與中國大陸平行，並被寬達 90 英里的最窄處所分隔開來。它從北北東伸展到南南西，最大長度為 243 英里，寬度一般在 60 到 80 英里之間，面積近 14,000 平方英里——大致相等於馬賽諸斯、羅德島和康乃狄克三州結合在一起。
Full frozen AMT lives in the private production tree (audiobook-albert-lai/01_AMT_scripts/zh-TW/AMT_v1.0/ch01.md), not on this public site.
Production notes
- SKU B runtime (full book body): ~4–5 h estimated (zh-TW often ~0.9–1.1× EN duration)
- Ch.I alone (sample): ~5–7 min across three voices
- Permission: commercial production GREEN (2026-07-09)
- Next after your review: lock zh-TW voice direction → full SKU B render with locked voice → multi-track hub
- GLOSSARY:
01_AMT_scripts/zh-TW/GLOSSARY.md— Formosa/Formosan policy, name bible