Skip to content

AI turns documents or topics into real, native PowerPoint decks—with native shapes, transitions and animations, data-backed charts and tables on demand, audio narration from speaker notes, and support for your own .pptx templates. · by Hugo He

docs(roadmap): log Confirm UI + source-fidelity gains, add beautification as next direction

hugohe3c1715d9f7adc
diff --git a/docs/roadmap.md b/docs/roadmap.md
index 0a7029bf..f0afd499 100644
--- a/docs/roadmap.md
+++ b/docs/roadmap.md
@@ -53,6 +53,8 @@ The past two months' structural capability growth. Single flags / incremental po
- **mode catalog grows to 5: add `briefing`** — fills the "neutral information delivery" cell: no thesis, no story, no teaching, no spectacle — topic titles, even weight, complete and scannable, for status updates / reference decks / catalogs / meeting packs / FAQs that inform without arguing. The five now partition presentation *intent* more nearly MECE: persuade (pyramid) · tell a story (narrative) · teach (instructional) · impress (showcase) · simply inform (briefing). `_index` gains a `briefing` vs `pyramid` tie-breaker ("if you'd have to invent a thesis to fit pyramid, it's briefing"). Five presets plus a `custom` escape hatch for a bespoke direction none of them captures (a special cadence, a multi-mode fusion, a particular posture) — either user-requested or Strategist-recommended, confirmed like every lock; a deck always locks one value, and a fusion is one `custom` describing the acts. The only thing to avoid is defaulting to `custom` when a preset genuinely fits. Mirrors the truth-precedence rule that a user-authored outline or direction overrides the mode
- **mode / visual-style system validated on real decks + four calibration tightenings landed** — the 5 modes + 18 visual-styles + the `custom` escape hatch were exercised on five covering decks (briefing×data-journalism / narrative×photo-editorial / instructional×chalkboard / showcase×glassmorphism / custom×zine, with the narrative deck going through the AI-image branch): **zero selection misses** (all four Close-calls tie-breakers fired and held against the real grey-zone pulls), **discipline fully honored** (zero HEX / Reference strength / whole-page layout language), **the `custom` mechanism works** (a `mode_behavior` prose paragraph carried a 10-page generation and reads as plain language at confirmation, not a bare token), **mode ⟂ visual_style holds** (any combination without bleed, including a positive check of the "keynote/launch = mode not style" route), export 5/5 decks × every page with 0 failures. Tightened four spots from the real signal: `strategist §e` now locks the full neutral tier set the visual_style implies up front (killing the mid-deck color top-ups seen in three consecutive decks), `executor-base §1` re-skins chart/layout template gradients-shadows-fills to the deck's visual_style (templates supply structure, not skin; mirror templates stay verbatim per §1.1), `briefing §1` makes `core_message` state coverage rather than a claim (briefing-only; global §IX assertion semantics stay for narrative/instructional/pyramid), and `svg_quality_checker` fixes font-family drift false positives (delimiter-matched capture + stack normalization) and drops the font-size ceiling for showcase mode and poster visual-styles
- **Optional spec-review checkpoint shipped** ([`workflows/refine-spec.md`](../skills/ppt-master/workflows/refine-spec.md)) — an opt-in pause after the Eight Confirmations: when the user explicitly asks (default OFF), the Strategist produces the full `design_spec.md` + `spec_lock.md`, then stops so the user can revise any part of the spec (outline / color / typography / layout / image strategy / page rhythm) before generation, keeping both files in sync. Same shape as the split-mode note — it never fires on its own, the default pipeline is unchanged, and it surfaces as one opt-in line inside the Eight Confirmations. Review lenses (logical clarity / information density / focus / register / emotional resonance / chapter balance + the design dimensions) give a direction only, never a numeric threshold (`Reference` strength). Prompted by [@cuberoocp](https://github.com/cuberoocp) in [issue #173](https://github.com/hugohe3/ppt-master/issues/173)
+- **Interactive visual Eight Confirmations page (Step 4)** ([`scripts/confirm_ui/server.py`](../skills/ppt-master/scripts/confirm_ui/server.py), field schema [`scripts/docs/confirm_ui.md`](../skills/ppt-master/scripts/docs/confirm_ui.md)) — the Eight Confirmations move from chat-only to a browser page that auto-launches by default at Step 4: enumerable fields (canvas / mode / visual_style / icons / image usage / formula & generation policy) list common options from `catalogs.json`, while generative fields present candidates — color swatches, live font previews (CJK / Latin previewed independently), and AI-image rendering × palette picks; plus live swatch feedback on custom HEX, a live combined color × typography preview, and a body-size range hinted from the chosen canvas. It **shares port `5050`** with the Step 6 Live Preview (they never run at once — the page auto-`--shutdown`s at the end of Step 4 to free the port; if the port is busy the launcher auto-advances to the next free one). On principle **chat stays the canonical channel and the page is a convenience layer**: the page's `result.json` is authoritative over the recommendations and is consumed in place downstream (image plan / `image_strategy` / fonts / the split and refine-spec toggles all read from it), and any failure to open / timeout / headless host degrades losslessly to the chat-summary fallback
+- **A batch of source-conversion fidelity gains** — less information dropped as source material enters the pipeline: `doc_to_md` converts Word OMML / Office Math equations to inline LaTeX, `pdf_to_md` recognizes `Figure N |` pipe-delimited captions, and `ppt_to_md` preserves hyperlinks already in the source deck (run-level external `[text](url)` / slide-internal jumps `[text](#slide-N)` / shape-level clicks, with dangerous-scheme filtering and anchor-text Markdown escaping). Caption recognition is based on [@suay1113](https://github.com/suay1113)'s [PR #191](https://github.com/hugohe3/ppt-master/pull/191); hyperlink preservation is distilled from [@ZhaoZuohong](https://github.com/ZhaoZuohong)'s [PR #155](https://github.com/hugohe3/ppt-master/pull/155)
---
@@ -60,7 +62,8 @@ The past two months' structural capability growth. Single flags / incremental po
Directions actively underway or up next, with no committed timeline.
-- No committed next capability right now. The mode / visual-style validation and calibration is settled (see "2026-06" above): the structure (5 modes + 18 visual-styles + custom) is locked, the four adjacent pairs are consolidated into one Close-calls table, and the four calibration tightenings have landed. Future direction is driven by real usage signal and feedback; long-running improvements are under "Ongoing maintenance" below, and evaluated-out directions under "Non-goals"
+- **Content-faithful PPT beautification / re-layout (next direction)** — a new route that mirrors `template-fill-pptx`: where `template-fill` reuses a deck's design and swaps in new content, this one inverts it — keep a deck's content, redo its layout. Given an existing PPTX, **every piece of text is preserved verbatim (nothing added / removed / reworded)**; beautification happens only in layout, hierarchy, whitespace, pagination, and visual structure (an ugly deck is usually ugly in its density and hierarchy, not in its colors or fonts). It also **extracts and inherits the source deck's visual identity (palette / fonts / layout idiom)** so the newly generated native elements match the original — they can be copied straight back into the original PPT and still look coherent. Because of that, this route does *not* freely apply the `visual-style` catalog; it carries over the style lifted from the source deck (the existing [`create-brand`](../skills/ppt-master/workflows/create-brand.md) / `pptx_to_svg` theme extraction already does this). Technically it is still "generate a native, editable PPTX from scratch" (`ppt_to_md` extracts content → main pipeline → a brand-new deck), not a patch over the original file, so it stays clear of the #53 "text-only fill of an arbitrary template" line under Non-goals. The contract: **text untouched; visual identity taken from the source deck, not freely reskinned; all the freedom to beautify goes to layout and hierarchy.** Likely shape: a standalone workflow plus a content-faithful Strategist sub-mode. Open boundaries still to settle: strict pagination fidelity vs. allowing re-pagination, and how to handle tables / charts / existing images
+- Otherwise: the mode / visual-style validation and calibration is settled (see "2026-06" above) — the structure (5 modes + 18 visual-styles + custom) is locked, the four adjacent pairs are consolidated into one Close-calls table, and the four calibration tightenings have landed. Apart from the direction above, future direction is driven by real usage signal and feedback; long-running improvements are under "Ongoing maintenance" below, and evaluated-out directions under "Non-goals"
---
diff --git a/docs/zh/roadmap.md b/docs/zh/roadmap.md
index f5e70d7f..ed462ef4 100644
--- a/docs/zh/roadmap.md
+++ b/docs/zh/roadmap.md
@@ -53,6 +53,8 @@
- **mode catalog 扩档至 5 个:加 `briefing`** — 补上「中性信息平铺」这一格:无论点 / 无故事 / 不教学 / 不冲击,topic 标题、等权铺事实、完整可扫读,服务周报 / 参考册 / 目录 / 会议材料 / FAQ 这类「只告知不论证」的 deck。五个 mode 自此更接近 MECE 地切分**表达意图**:说服(pyramid)· 讲故事(narrative)· 教会(instructional)· 震住(showcase)· 只告知(briefing)。`_index` 加了 `briefing` vs `pyramid` 的灰区判据(「要不要造个 thesis 才塞得进 pyramid → 那就是 briefing」)。五个预设之外加一个 `custom` 兜底,承接预设盖不住的 bespoke 方向(特殊节奏 / 多 mode 融合 / 特定姿态)——用户点名**或策略师推荐**皆可,与所有锁一样由用户确认;一份 deck 永远只锁一个值,融合=一个 custom 描述多幕。唯一要避免的是「预设明明贴合却图省事甩 custom」。这与「用户自带大纲 / 方向覆盖 mode」是同一条真值优先原则
- **mode / visual-style 体系真实 deck 验证完成 + 四项校准收紧落地** — 5 mode + 18 visual-style + `custom` 逃生舱在 5 份覆盖性 deck 上跑过验证(briefing×data-journalism / narrative×photo-editorial / instructional×chalkboard / showcase×glassmorphism / custom×zine,其中 narrative 一份走 AI 图生成分支):**选型零误判**(四对 Close-calls 灰区引力全被触发且全抗住)、**纪律全落实**(零 HEX / Reference 强度 / 整页版面语言)、**custom 机制可用**(`mode_behavior` 散文段落撑过 10 页生成、能讲成大白话让用户确认)、**mode ⟂ visual_style 正交成立**(任意组合无串味,含「keynote/发布会=mode 不是 style」路由正面验证)、导出 5/5 deck × 全页 0 失败。据真实信号收紧四处:`strategist §e` 按 visual_style 预判中性档位一次锁全(消除连续三份的 Executor 中途补色)、`executor-base §1` 套模板页重皮到当前 visual_style(模板供结构不供皮,镜像模板仍按 §1.1 逐字保留)、`briefing §1` 的 `core_message` = 本页覆盖什么而非证明什么(briefing 专属例外,全局 §IX 论断语义保留给 narrative/instructional/pyramid)、`svg_quality_checker` 修字体 drift 误报(按定界符匹配 + font-stack 归一化)+ 放宽 showcase mode 与 poster 类 visual-style 的字号上限
- **可选 spec 复核环节上线**([`workflows/refine-spec.md`](../../skills/ppt-master/workflows/refine-spec.md)) — 八项确认后新增一个 opt-in 停顿点:用户明确要求时(默认 OFF),Strategist 先产出完整 `design_spec.md` + `spec_lock.md`,停下来让用户对 spec 任意部分(大纲 / 配色 / 排版 / 版式 / 图片策略 / page rhythm)深入讨论修改,改完同步两个文件再进生成。与 split-mode 同构——不主动触发、默认管线零变化,仅在八项确认里多一行 opt-in 提示。复核视角(逻辑清晰度 / 信息密度 / 焦点 / 口语化 / 感染力 / 章节配比 + 各设计维度)只给方向、不落任何数字阈值(`Reference` 强度)。启发自 [@cuberoocp](https://github.com/cuberoocp) [issue #173](https://github.com/hugohe3/ppt-master/issues/173)
+- **交互式可视化八项确认页(Step 4)**([`scripts/confirm_ui/server.py`](../../skills/ppt-master/scripts/confirm_ui/server.py),字段 schema [`scripts/docs/confirm_ui.md`](../../skills/ppt-master/scripts/docs/confirm_ui.md)) — 八项确认从纯聊天升级为浏览器可视化页面,Step 4 默认自动拉起:可枚举字段(canvas / mode / visual_style / 图标 / 图片用法 / 公式与生成策略)从 `catalogs.json` 出常用项、生成型字段摆候选——配色 swatch、字体实时预览(CJK / Latin 各自独立)、AI 图 rendering × palette 候选;并支持自定义 HEX 即时 swatch 反馈、配色 × 字体组合的实时合成预览、canvas 尺寸联动建议正文字号区间。与 Step 6 Live Preview **共用 `5050` 端口**(永不同时跑,确认页在 Step 4 末尾自动 `--shutdown` 让出端口;端口被占则自动顺延到下一个空闲口)。原则上**聊天始终是 canonical 通道、确认页只是便利层**:页面写出的 `result.json` 对推荐值具权威性、由下游就地消费(图片计划 / `image_strategy` / 字体 / split 与 refine-spec 开关都从中读取),任何打不开 / 超时 / 无 GUI 的情况都无损回退到聊天总结路径
+- **源文档转换保真度提升一批** — 源材料进管线时更少丢信息:`doc_to_md` 把 Word 里的 OMML / Office Math 公式转成内联 LaTeX、`pdf_to_md` 识别 `Figure N |` 竖线分隔的图注、`ppt_to_md` 保留源 deck 已有的超链接(run 级外链 `[text](url)` / slide 内部跳转 `[text](#slide-N)` / shape 级点击,含危险 scheme 过滤与锚文本 Markdown 转义)。图注识别基于 [@suay1113](https://github.com/suay1113) [PR #191](https://github.com/hugohe3/ppt-master/pull/191),超链接保留提炼自 [@ZhaoZuohong](https://github.com/ZhaoZuohong) [PR #155](https://github.com/hugohe3/ppt-master/pull/155)
---
@@ -60,7 +62,8 @@
明确在做或下一步要做的方向,不承诺时间窗口。
-- 暂无已承诺的下一步能力方向。mode / visual-style 体系的验证与校准已收口(见上「2026-06」),结构(5 mode + 18 visual-style + custom)定型、四对近邻消歧并成一张 Close-calls 表、四项校准收紧已落地。后续方向由真实使用信号与反馈驱动;长期改进见下「持续维护方向」,已评估不做的见「明确不做」
+- **内容保真的 PPT 美化 / 重排版(下一步方向)** — 一条与 `template-fill-pptx` 互为镜像的新路线:`template-fill` 是「复用某份 deck 的设计、换上新内容」,这条反过来——「保留某份 deck 的内容、重做它的版面」。输入一份现成 PPTX,**逐字保留全部文字内容(不增 / 不删 / 不改写)**,美化只发生在版面、层级、留白、分页与视觉结构上(丑 deck 的丑通常在密度与层级,不在它的色与字)。同时**从原 deck 提取并继承其视觉身份(配色 / 字体 / 版面语汇)**,让新生成的原生元素与原 deck 风格一致——可以直接复制回原 PPT 模板、看上去仍然协调;也因此这条路线不自由套用 `visual-style` catalog,而是沿用从源 deck 抽取的风格(现有 [`create-brand`](../../skills/ppt-master/workflows/create-brand.md) / `pptx_to_svg` 主题提取已能做)。技术上仍是「从零生成原生可编辑 PPTX」(`ppt_to_md` 抽内容 → 主管线 → 全新 deck),不是在原文件上打补丁,因此不碰 Non-goals 里 #53 那条「读任意模板只填字」的红线。关键契约:**文字一字不动;视觉身份取自原 deck,不自由换皮;美化的自由度全部留给版面与层级**。落地形态预计为一条独立 workflow + Strategist 的内容保真子模式。仍待定的边界:严格分页保真 vs 允许重排分页、表格 / 图表 / 已有配图如何处理
+- 其余:mode / visual-style 体系的验证与校准已收口(见上「2026-06」),结构(5 mode + 18 visual-style + custom)定型、四对近邻消歧并成一张 Close-calls 表、四项校准收紧已落地。除上面这条外,后续方向由真实使用信号与反馈驱动;长期改进见下「持续维护方向」,已评估不做的见「明确不做」
---