From 2b0411bcb0139d2341cf5f8e558057eb21603288 Mon Sep 17 00:00:00 2001 From: zhanglinghao Date: Wed, 7 Oct 2026 21:48:30 +0800 Subject: [PATCH 1/3] Lessons from a narrated film with 3D: one voice in a one-request take, labels at the floor from the start, text swaps and readouts, glass in Three.js MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The maintainer asked that the general lessons of a 2-minute narrated canvas + Three.js film go into the docs as short guidance. - playbook/04: in a one-request gemini take every line still carries its own style, so a per-line [direction] makes that line sound different; for one voice, no per-line directions. Evening out the pace line by line on one take (waveform-refined cuts, capped inner pauses, partial per-line atempo towards one target), described, not a command. When transcription is unavailable: the take is on disk; --resume later or align locally with whisper, mapping digits back to spelled-out years. - tools/audio/README: a daily-quota 429 was once waited on and retried like the per-minute limit, which looked hung. - playbook/03 §4: write labels at the floor from the first draft; a shot's headline is a 主标题; credits go into the description. - playbook/02 layer 3, TASTE_CHECKLIST #19: strips at text swaps and rolling readouts; a readout rests only on sourced values. - engines/README: RoomEnvironment reflections and the default LatheGeometry seam on glass; a fresnel rim shell instead. --- CHANGELOG.md | 8 ++++++++ engines/README.md | 1 + playbook/02-verification.md | 1 + playbook/03-motion-design.md | 1 + playbook/04-audio.md | 11 +++++++++++ templates/TASTE_CHECKLIST.md | 2 +- tools/audio/README.md | 2 +- 7 files changed, 24 insertions(+), 2 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index a662bfc..01b682e 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -2,6 +2,14 @@ ## Unreleased +**Lessons from a narrated film with 3D: one voice across a one-request take, labels sized from the start, text swaps and readouts, glass in Three.js** +- Why: the maintainer asked that the general lessons from a 2-minute narrated canvas + Three.js film go into the docs as short guidance. +- playbook/04: in a one-request gemini take (`--join all`) every line still carries its own style, so one line with a `[direction]` (often the opening) sounds different from the rest; for one voice, put the whole delivery in the overall style and give no line a direction. How to even out the pace line by line on one take without synthesizing again (cuts refined from the waveform, so a misheard first syllable is not clipped; long pauses inside a line capped; per-line `atempo` part of the way towards one target pace), described, not a command yet. When transcription is unavailable (daily quota, API errors): the take is already on disk; `--resume` later, or align locally with whisper, mapping its digits back to a script that spells years out (一九五三). +- tools/audio/README: a daily-quota 429 is not always refused at once; one was waited on and retried like the per-minute limit (up to 5 times a call), which looked hung. +- playbook/03 §4: write readable labels at the floor from the first draft (44 px for `desktop`; one draft had about 40 at 24–40 px); a shot's one-line headline counts as a 主标题 (84 px); source and credit lines go into the description. +- playbook/02 layer 3 and TASTE_CHECKLIST #19: a strip at each text swap in one place (old out before new in; a reviewer found three that the author's own check missed) and at each rolling readout, whose resting values must be sourced, not interpolated. +- engines/README (Three.js): `RoomEnvironment` reflections drew a bright vertical strip down every glass object, and a full-turn `LatheGeometry` seam faces a +z camera by default; turn the seam away and give glass a fresnel rim shell instead of the environment map. + **`bin/vh tts … minimax`: MiniMax narration in one request, timed by its own subtitles** - Why: a film's narration was made with MiniMax T2A (`speech-2.8-hd`) through a project script. Sending the whole script in one request kept the voice from drifting between blocks, which the maintainer had heard and disliked with block-by-block synthesis, and the request's own subtitles gave every character's time without an ASR pass. This makes that a provider. - `minimax` posts to `https://api.minimax.io/v1/t2a_v2` (the China host `api.minimaxi.com` rejects international keys with 2049; `MINIMAX_BASE_URL` for a China-platform key). Key from `MINIMAX_TOKEN_PLAN_KEY` (Token Plan, spends its credits) first, then `MINIMAX_API_KEY` (pay-as-you-go); an error names the variable, never the key, with a hint for 1008 (no balance on that billing) and 1004/2049 (wrong host for the key). `MINIMAX_TTS_MODEL` / `_SPEED` / `_VOL` / `_PITCH` are recorded in `_run.json`, so `--resume` refuses a change; speed, volume and pitch are checked against the API's ranges before anything is deleted, and so is the 10,000-character limit of each request. `--instruct` / `[direction]` count only as one of MiniMax's nine emotion words; other text is ignored and that is said once. Default voices: zh `Chinese (Mandarin)_Reliable_Executive`, en `English_expressive_narrator`. diff --git a/engines/README.md b/engines/README.md index 7d43806..b96b00d 100644 --- a/engines/README.md +++ b/engines/README.md @@ -161,6 +161,7 @@ window.__vh = { - **把 3D 点投影到屏幕前先 `camera.updateMatrixWorld()`**：同一帧里刚 `lookAt` 过，矩阵还没更新，钉在 3D 点上的标注会错位。 - **`EdgesGeometry(geo, 阈值角)` 只画相邻面夹角大于阈值的边**：20 段的圆柱相邻面差 18°，阈值设 20° 就一条侧边线都画不出来。线框画法的圆柱少分几段（12 段是 30°），或者把阈值调低。 - **物体的位置在建的时候就设好**：只在每帧的动画函数里设 `position` 的话，不参与动画的零件永远停在原点（盒子以中心为原点，会半截埋进地面）。 + - **玻璃器皿上一道从上到下的亮条**：用 `RoomEnvironment` 做环境贴图时，它的反射会在每个玻璃件上留下一道亮竖条，一排烧瓶就是一排亮条；`LatheGeometry` 转满一圈的接缝默认也正对 +z 方向的镜头，一起转到背面（`phiStart` 设 `Math.PI`）。玻璃所在的场景不设 `scene.environment`，玻璃感改由一层菲涅尔边光壳来给：同一个几何体放大约 1.004 倍，`ShaderMaterial` 加法混合、不写深度，亮度取 `pow(1 − |N·V|, 2.6)`，只有边缘微微发亮。 - **渲染时不要依赖在线 CDN。** importmap 指向 CDN 时，只要页面加载那一刻断网，渲染就会静默挂住：日志 0 字节，也不报错。介绍片的第一次 final 渲染就卡在这里。把 three.js 装进项目，importmap 指向 `node_modules`，snapshot 和 render 都能读到这些文件： ```bash diff --git a/playbook/02-verification.md b/playbook/02-verification.md index c855e9d..b291fbc 100644 --- a/playbook/02-verification.md +++ b/playbook/02-verification.md @@ -24,6 +24,7 @@ - **strip**：看时序。按固定步长 0.1–0.15 秒出一串帧，从头到尾像观众那样读，数一数每个 read 分到几帧（24 帧 = 1 秒）。 - **crop**：局部放大，看脸、手、接触点、发光区域。 - 另外，每个转场前后各 0.5 秒都要单独看；整片再按目标屏的尺寸看一遍（BRIEF 的 `Watch on`）；循环播放的交付物还要看末帧接首帧（两条命令见下文"手机测试与循环接缝"）。 + - 同一位置换字、会滚动的读数，各看一条 strip：换字时旧字退完新字才进（#17；一支 2026-10 的片子作者自查漏了三处，是 reviewer 看出来的）；读数每次停下的数都和 NOTES 里的出处对一遍，停在插值算出来的数上算编造（#19）。 4. **VLM 批评要给离散选项**：让模型从锚点网格里选位置（Code2Video 的做法），或者从几个渲染出来的变体里挑一个（Paper2Video 的做法）。不要让它输出"往左移 12px"这种数值微调。每轮最多列 3 个问题，每个问题都附一条可以直接执行的修改。 5. **reviewer subagent**：开一个全新上下文（`standard` 用便宜一档的模型，Claude Code 里 `model: sonnet`；`studio` 可以用最强档），只给它联系表、`TASTE_CHECKLIST.md` 和 `STORYBOARD.md`，让它逐条打分；不合格的退回给写代码的 agent。 - **人设**：严苛的导演，不是为作品自豪的作者。提示词的大意是："这支片子不是你做的。你是一位挑剔的导演，任务是找出它为什么还不能发。"作者自己看久了会习以为常，换一个没有感情投入的人才看得出。 diff --git a/playbook/03-motion-design.md b/playbook/03-motion-design.md index 2d27ed5..d68ba3f 100644 --- a/playbook/03-motion-design.md +++ b/playbook/03-motion-design.md @@ -134,6 +134,7 @@ - `feed` 那一档的字要大得多，因为横屏片在信息流里本来就小。能另出一个竖版（重新构图，不是裁）就另出，比把横屏的字放大到这一档好。 - **4K**：照常按 1080p 写合成和字号，出片时渲成 4K（HyperFrames `render --resolution 4k`，竖屏 `portrait-4k`，方形 `square-4k`；见 `engines/README.md`），画面和字一起放大一倍，屏幕上看起来一样大。电脑上看 4K，字不需要占画面更大的比例；只有直接按 3840 宽写合成时，表里的 px 才要乘 2。 - 风格预设的 STYLE.md 里写的字号，是它 5 s 样片用过的数（不少标签是 30–42 px），不是下限；借进片子以后，标题、标签、字幕按这张表来。 + - **写场景时就从下限写起**：标签、注释、读数这些要读的辅助字，一开始就按这一档的下限写（`desktop` 是 44 px），不要先写小、等自查再放大：一支 2026-10 的横屏片初版有约 40 处辅助字是 24–40 px，事后放大要挨个重排。镜头里点题的那行大字（一句话标题）算主标题，不算辅助字（`desktop` 是 84 px）。出处、图片署名、素材来源这类不用观众读的小字，放进视频简介，不留在画面上。 - 来源：`docs/research/06-concept-first-ab.md`（同一个 44 px，竖屏手机上约 15 px，横屏放进信息流只有约 8 px）；屏幕宽度是常见设备的 CSS 宽度【综合】。 - **中文字体**【综合】：思源黑体/Noto Sans CJK 用于正文和字幕，思源宋体用于标题和人文类内容；使用前确认本机已安装，或者把字体文件放进项目的 `assets/`。 - **decode / scramble 字**：字像被程序一点点解码出来，是代码感片子的常用效果。守三条规矩： diff --git a/playbook/04-audio.md b/playbook/04-audio.md index 6135edc..850df7c 100644 --- a/playbook/04-audio.md +++ b/playbook/04-audio.md @@ -70,6 +70,8 @@ bin/vh beats <任意音乐文件> # 外来音乐的节 `--align gemini` 对任何 provider 都能用：每句合成完交给 Gemini 转写，拿回带时间戳的词，再和稿子逐字对齐。结果写进 timeline：`words` 是逐词时间（文字用稿子的写法），字幕和画面逐词同步用它；`asr` 是和稿子的相似度，跑偏、或者句首句尾多出 1 s 以上声音的句子标上 `flag`，命令最后逐句列出来。转写失败也不丢配音：修好原因后，同一条命令加 `--resume` 接着跑。时间步长、费用、限额和实测见 `tools/audio/README.md` 的同名一节。 +**转写用不了时**（当天额度用完，或者接口一直报错）：额度用完后，`--align gemini` 可能在 429 上一遍遍等待重试，看着像卡住。合成好的音频已经在盘上（逐句合成时是 `voiceover..wav`，`--join` 时是 `vo//_blockNN.wav`），停掉不会丢；可以等额度恢复后加 `--resume` 接着转写，也可以改用本地 whisper 拿逐字时间（还没封装；Apple Silicon 上用 mlx-whisper 的 `whisper-large-v3-turbo`，`initial_prompt` 里写上人名和术语），再按字和稿子对齐（difflib；没对上的字在前后两个对上的字之间插值）。whisper 把年份写成阿拉伯数字：稿子为了念法写成"一九五三"时，对齐前先把这些数字换回稿子的写法，否则那句的起点会算晚，剪的时候切掉第一个字。 + 有词级时间时，`bin/vh captions` 会给每条字幕加上 `words`，念得太快的短字幕会往后延到 1.8 s（`tools/audio/README.md` 的"字幕"）。 **双人对话、整段合成、音色、费用**：`@speakers` 双人对话、`--join` 整段合成、音色库和设计音色（`bin/vh voices`）、Gemini 的费用和限额，都在 `tools/audio/README.md` 的同名各节。用了 Gemini 旁白的片子，在 `NOTES.md` 写明"旁白为 AI 合成（Gemini TTS，含 SynthID 水印）"，发布时按平台要求标注。 @@ -188,6 +190,7 @@ bin/vh qa mix $A/stems --beats $A/music.beats.json # 只看混音报告（词 - 同一段里相邻两句的语气要有落差：问句接答句，铺垫接爆点，快接慢。 - **标签**：``、``、``、``、`` 可以直接写进句子里。只有 `gemini` 会演出来，其他 provider 和字幕都会自动去掉（规则见 `tools/audio/README.md`）。`minimax` 用的是停顿标记 `<#0.4#>`（秒），同样只有它会停。中文稿里也写英文标签，官方说这样效果最好。标签只管某一刻的动作（停顿、呼吸、笑），持续的语气写在 `[ ]` 里。 - **给 gemini 的指示要短**：一句话讲清情绪和节奏就够了。年龄、性别、口音属于音色，不要写进指示；长段的人设和导演笔记容易让音色漂移。 +- **整篇要一个声音时，不写逐句指示**：gemini 整稿一次合成（`--join all`）时，每句仍是单独的一段文字，各带自己的 style：整体语气，加上这句 `[ ]` 里的指示。只有一句带 `[ ]`（常见的是开头那句），这句就会和其余的听起来不一样。要全片一个声音，把想要的语气都写进整体语气，每句都不写 `[ ]`。 - **谁能演**： - `gemini` 表演力最好，整体和逐句指示都听； - `minimax` 只认一个情绪词（`calm`、`happy`、`whisper` 等），写别的会被忽略； @@ -215,6 +218,14 @@ bin/vh qa mix $A/stems --beats $A/music.beats.json # 只看混音报告（词 **语速的依据和用法**：表里是句内语速，不算句间停顿。普通话正常朗读、去掉 150 ms 以上的停顿后约 5.2 字/秒：七种语言的朗读语料里，汉语是 5.18 音节/秒，一个汉字一个音节（Pellegrino, Coupé & Marsico 2011，*Language* 87(3)，[doi:10.1353/lan.2011.0057](https://doi.org/10.1353/lan.2011.0057)）。电话闲聊按总时长算是每分钟 228–247 字，约 3.8–4.1 字/秒（Yuan, Liberman & Cieri 2006，Interspeech，[doi:10.21437/interspeech.2006-204](https://doi.org/10.21437/interspeech.2006-204)）。知识科普的 4.5–5.5 围着正常朗读，原理讲解慢一档。倒推稿子字数时乘 0.85，留出句间停顿【综合】：45 s 的知识短视频约 170–210 字，3 分钟的讲解约 540–690 字。TTS 的实际语速看音色：本节第 4 小节比较 `duck_ratio` 用的那段测试旁白（0.25 s 间距，本地 Qwen3-TTS 0.6B 的 Serena，7 句 107 字）句内约 4.1 字/秒。所以字数只用来起稿，时长以 `timeline.json` 的实测为准（硬规则 2）。 +**一条 take 里快慢不一时，逐句拉齐**：整稿一次合成音色不漂，句与句的语速却可能差不少（一条 Gemini 整稿实测 5.0–7.8 字/秒）。不用重合成，在这条 take 上逐句处理就行（还没做成命令）： +1. 起止按波形定：从对齐给的句首往回走，走到声音静下来为止（ASR 听错第一个字时给的起点偏晚，照着剪会切掉它），句尾同样往后走； +2. 句内过长的停顿压到约 0.25 s，切口两侧各淡几毫秒； +3. 每句用 `atempo`（不变调）拉向同一个目标语速（取人认可的那一段，比如开头），只校正一部分（例如差距的 0.8 次方，倍数限在 0.92–1.18），留一点自然的起伏； +4. 重排句间停顿，整条再对齐一次，拿最终的逐字时间。 + +一支 2026-10 的片子这样拉到 5.92 ± 0.11 字/秒，听起来还是同一条 take。 + **要做 A/B**：`studio` 档位下，同一段旁白至少试两种导演方向，把两版都给人听。实测对比见本节开头：同样三句话，只给一个"纪录片旁白"的整体语气时听起来平；逐句导演、再卡上拍之后，才有起伏和节奏。 ### 2. 旁白卡拍（动效类必做） diff --git a/templates/TASTE_CHECKLIST.md b/templates/TASTE_CHECKLIST.md index 5558653..b262825 100644 --- a/templates/TASTE_CHECKLIST.md +++ b/templates/TASTE_CHECKLIST.md @@ -43,7 +43,7 @@ 18. 超过 3s（短视频）或 8s（讲解）什么都没发生？停拍时声音和镜头同时停死（观众会以为播放卡住了；#16 的停顿要像屏息：低音垫着、镜头慢漂）？【底线】分镜里声明的卡点没对准节拍（±1 帧）？片中（首尾之外）出现数字静音或近乎静音（有声片才查；用户要的无声片不带音轨）？ **内容** -19. 第 1 秒看不出这是什么、为什么要看？【底线】有占位文字、编造的数字？把不存在的界面当成真实产品的界面展示（立意本身就是一个明显虚构的界面时不算，例如把讲解装进一个仿真的 app）？ +19. 第 1 秒看不出这是什么、为什么要看？【底线】有占位文字、编造的数字（在几个有出处的数之间滚动的读数，停住让人读的只能是有出处的数，不能停在插值算出来的值上）？把不存在的界面当成真实产品的界面展示（立意本身就是一个明显虚构的界面时不算，例如把讲解装进一个仿真的 app）？ 20. 把这一帧换到任何别的主题也照样成立（说明它是装饰，不是设计；最常见的通用 AI 装饰是四角的小标签、取景框式的边角线、不带信息的 HUD 读数）？说不出这一镜看完观众多知道了什么？ ## 结构化输出 diff --git a/tools/audio/README.md b/tools/audio/README.md index 65025c0..da55bdb 100644 --- a/tools/audio/README.md +++ b/tools/audio/README.md @@ -95,7 +95,7 @@ - **费用**（Standard 档，每百万 token）：Flash TTS 输入 $0.50、输出音频 $9（约 $0.00225 / 10 s）；Flash-Lite TTS 输入 $0.50、输出 $6（约 $0.0015 / 10 s）。这是 2026-12-31 之前的价格，**2027-01-01 起全部翻倍**（Flash $1 / $18，Lite $1 / $12）。Batch 和 Flex 档是一半。免费档不收费，但内容可能被用来改进 Google 的产品，有保密要求的稿子用付费档； - 单个请求最多 8,192 个输入 token； -- **转写的配额**（`--align gemini`，2026-10-01 在 Tier 1 上）：`gemini-3.5-transcribe` 每分钟 10 次、每天 100 次，官方文档说按项目算，不按 key。每分钟的 429 会自动等；每天的用完会直接失败，额度恢复后用 `--resume` 接着跑，已经转写成功的句子（`vo//*.asr.json`）不会重做。每天几点恢复没有定论：文档写太平洋时间零点，429 提示里的倒计时指向 UTC 零点。AI Studio 的用量页显示的是 28 天的峰值，不是今天还剩多少； +- **转写的配额**（`--align gemini`，2026-10-01 在 Tier 1 上）：`gemini-3.5-transcribe` 每分钟 10 次、每天 100 次，官方文档说按项目算，不按 key。每分钟的 429 会自动等；每天的用完时，429 要求等 90 s 以上就直接失败；但也遇到过日配额用完后它照样等了再试（那时 429 要的等待不到 90 s，或者没写、按 60 s 算），每次调用最多 5 次，看着像卡住，怎么办见 `playbook/04-audio.md` 的"转写用不了时"。额度恢复后用 `--resume` 接着跑，已经转写成功的句子（`vo//*.asr.json`）不会重做。每天几点恢复没有定论：文档写太平洋时间零点，429 提示里的倒计时指向 UTC 零点。AI Studio 的用量页显示的是 28 天的峰值，不是今天还剩多少； - **语气写短**：`--instruct` 和 `[指示]` 最后都进 `speech_metadata.style`。官方建议只写这一句的情境语气（"压低声音，卖个关子"）；年龄、性别、口音这类身份特征不要写进 style，要换音色，或者用 `voices design` 做一个。长篇的"角色设定""导演笔记"是音色漂移最常见的原因； - **每段 Gemini 音频都带 SynthID 水印**（听不出来，但能检测到）。用了 Gemini 旁白的片子，在项目 `NOTES.md` 里写明"旁白为 AI 合成（Gemini TTS，含 SynthID 水印）"，发布时按平台要求标注。 From a64f0a09426fe5813ef05666562f0e39fb150004 Mon Sep 17 00:00:00 2001 From: zhanglinghao Date: Wed, 7 Oct 2026 22:10:08 +0800 Subject: [PATCH 2/3] Fixes from the independent review MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - playbook/04: whisper output goes through tts.py's align_lines (it spreads spelled-out years over a heard "1953"; a home-made character aligner can misplace the line start); the one-voice advice is for single-narrator films and asks for one short overall style; the pace numbers are re-measured on the take with the same method (4.7–6.7 → 5.6–6.1 chars/s, sd 0.45 → 0.11); the evening steps say how the factor is computed, avoid anullsrc + concat, and write the take, the timeline and the captions back. - tools/audio/README: the quota text describes what the code does (a 429 under 90 s, or without a delay, is retried up to 5 times) and points to the playbook; the joined-synthesis section points to the one-voice advice. - playbook/03 §4: credits stay on screen where a licence or the type doc wants them (a data story's source line), as labels at the floor; the headline is "the line used as the shot's title". - TASTE_CHECKLIST #6: a shot's headline counts as a 主标题. - engines/README: "a row of glass objects"; dropping scene.environment also drops it for the other materials in that scene; the list's lead-in says most pitfalls come from the intro film. - playbook/02: "旧句退完、新句才进", as in #17. - CHANGELOG: rewritten to match, with the two floor changes said explicitly. --- CHANGELOG.md | 11 ++++++----- engines/README.md | 4 ++-- playbook/02-verification.md | 2 +- playbook/03-motion-design.md | 2 +- playbook/04-audio.md | 14 +++++++------- templates/TASTE_CHECKLIST.md | 2 +- tools/audio/README.md | 6 +++--- 7 files changed, 21 insertions(+), 20 deletions(-) diff --git a/CHANGELOG.md b/CHANGELOG.md index 01b682e..ba15b72 100644 --- a/CHANGELOG.md +++ b/CHANGELOG.md @@ -4,11 +4,12 @@ **Lessons from a narrated film with 3D: one voice across a one-request take, labels sized from the start, text swaps and readouts, glass in Three.js** - Why: the maintainer asked that the general lessons from a 2-minute narrated canvas + Three.js film go into the docs as short guidance. -- playbook/04: in a one-request gemini take (`--join all`) every line still carries its own style, so one line with a `[direction]` (often the opening) sounds different from the rest; for one voice, put the whole delivery in the overall style and give no line a direction. How to even out the pace line by line on one take without synthesizing again (cuts refined from the waveform, so a misheard first syllable is not clipped; long pauses inside a line capped; per-line `atempo` part of the way towards one target pace), described, not a command yet. When transcription is unavailable (daily quota, API errors): the take is already on disk; `--resume` later, or align locally with whisper, mapping its digits back to a script that spells years out (一九五三). -- tools/audio/README: a daily-quota 429 is not always refused at once; one was waited on and retried like the per-minute limit (up to 5 times a call), which looked hung. -- playbook/03 §4: write readable labels at the floor from the first draft (44 px for `desktop`; one draft had about 40 at 24–40 px); a shot's one-line headline counts as a 主标题 (84 px); source and credit lines go into the description. -- playbook/02 layer 3 and TASTE_CHECKLIST #19: a strip at each text swap in one place (old out before new in; a reviewer found three that the author's own check missed) and at each rolling readout, whose resting values must be sourced, not interpolated. -- engines/README (Three.js): `RoomEnvironment` reflections drew a bright vertical strip down every glass object, and a full-turn `LatheGeometry` seam faces a +z camera by default; turn the seam away and give glass a fresnel rim shell instead of the environment map. +- playbook/04: in a one-request gemini take (`--join all`) every line still carries its own style, so one line with a `[direction]` (often the opening) tends to sound different from the rest; for one voice in a single-narrator film, write one short overall style and no per-line directions (tools/audio/README's "整段合成" points there). How to even out the pace line by line on one take without synthesizing again (cuts refined from the waveform, so a first syllable the ASR placed late is not clipped; long pauses inside a line capped; per-line `atempo` part of the way towards one target pace; the take and the timeline written back, aligned again, captions redone), described, not a command yet: the film went from 4.7–6.7 to 5.6–6.1 CJK chars/s per line. When transcription is unavailable (daily quota, API errors): the take is already on disk; `--resume` later, or take word times from a local whisper and align them with tts.py's `align_lines`, mapping digits back to a script that spells years out (一九五三). +- tools/audio/README: a 429 asking for under 90 s (or naming no delay) is retried like the per-minute limit, up to 5 times a call, so a spent daily quota can look hung; the old text said the daily quota always fails at once. +- playbook/03 §4: write readable labels at the floor from the first draft (44 px for `desktop`; one draft had about 40 at 24–40 px). Source, credit and asset lines go into the description unless a licence or the type doc wants them on screen, and then they are labels. +- playbook/02 layer 3: a strip at each text swap in one place (#17's old-out-before-new-in, unchanged; a reviewer found three that the author's own check missed) and at each rolling readout, whose resting values are checked against NOTES. +- Floors, said explicitly (CONTRIBUTING rule 7): TASTE_CHECKLIST #19 names a readout resting on an interpolated value as a fabricated number (the film's reviewer had judged it under #19); #6 and playbook/03 §4 count a shot's one-line headline as a 主标题, so it is held to 84 px (150 for `feed`), not 44. +- engines/README (Three.js): `RoomEnvironment` reflections drew a bright vertical strip down every glass object, and a full-turn `LatheGeometry` seam faces a +z camera by default; turn the seam away and give glass a fresnel rim shell instead of the environment map. The list's lead-in now says most of its pitfalls come from the intro film. **`bin/vh tts … minimax`: MiniMax narration in one request, timed by its own subtitles** - Why: a film's narration was made with MiniMax T2A (`speech-2.8-hd`) through a project script. Sending the whole script in one request kept the voice from drifting between blocks, which the maintainer had heard and disliked with block-by-block synthesis, and the request's own subtitles gave every character's time without an ASR pass. This makes that a provider. diff --git a/engines/README.md b/engines/README.md index b96b00d..fb4e78a 100644 --- a/engines/README.md +++ b/engines/README.md @@ -152,7 +152,7 @@ window.__vh = { - **等宽字体的坑**：`ui-monospace, monospace` 能通过字体 lint，快照里看起来也是等宽，但 `hyperframes render` 渲染出来却是比例字体（showcase 00 在第一版全片都中了招）。要用 `@font-face` 显式声明，例如 `@font-face{font-family:"LF Mono";src:local("SF Mono"),local("Menlo")}`。**判断字体要看 mp4 渲染出的帧，不要看快照。**`check --snapshots` 只保存对比度检查用的 PNG，不能替代 `snapshot --at`。 - `snapshot` 在 tween 刚开始的那一刻，可能和最终渲染出的帧不一致。关键帧以渲染出的 mp4 为准，逐帧 strip 的做法见 `playbook/02-verification.md`。 - 完整样板：`showcase/02-short-leo-doppler/`（竖屏科普，单个 `index.html`）。 -- **HyperFrames 加 Three.js（3D 世界、一镜到底）的坑**，来自介绍片的制作： +- **HyperFrames 加 Three.js（3D 世界、一镜到底）的坑**，多数来自介绍片的制作： - 每一帧都用 `renderAt(t)` 从 t 算出摄像机和所有物体的状态。摄像机路线用按 t 参数化的样条，不用 Three.js 的动画时钟。关卡处可以用 stop 关键帧停站，但要叠一层低幅的手持漂移，镜头不要完全停死（原因见 `playbook/08-vfx-and-motion-sources.md` 的"一镜到底"一节）。 - `VideoTexture` 必须**每帧**设置 `texture.needsUpdate = true`，否则渲染出的屏幕是全黑的。视频按 t 去 seek，而且要等 seek 完成。 - `__hf.buildReady` 要在普通 `