Full setup (real engines)

```bash

1. clone + install

git clone https://github.com/ChrisChen667788/wind-comic.git cd wind-comic npm install

🏁 Quick start

🆕 Earlier: v6 → v10 — from *demo* to *production platform*

v3 shipped the pipeline. v6 turned it into a production studio; v7–v9 hardened it into a platform; v10 closed the lip-sync, template-market, and cost loops. Reusable characters, a prompt IDE, novel→season auto-splitting with real voiceover, a 60-style gallery, a director's control room, team credit budgets, an industry-grade script audit (Polish Pro, v7.1), a premium design pass (v8.3), a fully-migrated Postgres backend (v9), a lip-sync delivery pipeline + template market + cost observability (v10), and a live API health board — every screen below is a real capture of the running app.

<p align="center"> <img src="assets/v10/landing.jpg" alt="Wind Comic landing — 青枫漫剧 AI Animation Agent Studio, 8 agents · 7 engines · 3 consistency guards, looping cinematic hero" width="100%" /> <br/><sub>v10 landing — looping cinematic hero, 8 collaborating agents · 7 media engines · 3 consistency guards.</sub> </p>

🎨 Style Gallery — 60 cinematic looks, one click *(v6.3)*

Lock a consistent visual identity before you generate. Search, filter by category, and apply any preset straight into the creation workshop. <p align="center"><img src="assets/v6/styles.jpg" width="100%" /></p>

🎬 Screenshots

📸 v12.425 资产真机截图(captured 2026-09-06)

全部由 node scripts/capture-v12-425.mjs 对着本机真实数据跑出来。脚本带空壳检测与近重复检测: 页面没渲染出来、或者「点了 tab 但视图没变」的,一律删掉并如实报告 —— 一张截图冒充两张,比少一张更糟。 📚 完整 18 张 + 每张的实现逻辑与工作流架构说明:docs/SCREENSHOTS-v12.425.md (单独成文是为了不让首页替访客扛下载量 —— README 引用媒体有 12MB 预算门禁。) v12.426 补记:上面「我的项目」那张(docs/screenshots/v12/01-my-projects-manage.jpg)已重拍。 旧图里卡片大半是渐变占位 —— 不是没作品,是封面被冻结成了一次失败的快照:流水线末尾抄一份第 1 镜的 imageUrl 就再不重算,于是出图全挂时的 mock 兜底图被永久写死(30 个项目里 12 个如此,其中 3 个明明各有 11~12 张真分镜还活着)。现改为读时解析,并清掉了 21 个夹具/无素材项目(下架,可一键恢复)。

---

📸 v12.402 – v12.416 screenshots (captured 2026-09-03)

All captured by node scripts/capture-v12-416.mjs against real local data (1440×900 @2x, downscaled to 1600px wide). The script has blank-page detection and near-duplicate detection: a screen that failed to render, or one where "the tab was clicked but the view never changed", is deleted and reported honestly — one screenshot masquerading as two is worse than one missing.

Only one project-page shot: clicking the tabs and scrolling the body both leave the view unchanged. This conclusion was wrong and is retracted (v12.425.) The tab bar sits below the fold, so a coordinate click missed it; scrollIntoView followed by el.click() switches tabs fine. See the v12.425 section above for 12 shots taken that way.

Measured performance (local dev server, curl time-to-first-byte): landing 46 ms; console pages (projects / assets / engine health / usage) 22–38 ms. Tests: 5294 passing · 0 failing; preflight 10/10; CI all five jobs green.

Below is the foundational v3 pipeline (the v6 studio screens are in the New in v6 section above). Every panel is a real puppeteer capture of the running app (run node scripts/capture-screenshots.mjs / node scripts/capture-v6.mjs to refresh).

🔀 Gateway routing — default model map *(matches current `lib/config.ts`)*

Every model call is provider-pluggable (priority chain + automatic fallback). Creative and high-frequency LLM traffic are split across two model tiers, and MiniMax is always the last-resort fallback on any error / out-of-credits / timeout:

• Why two LLM tiers: the creative tier (DeepSeek -pro, a reasoning model) carries writer/director quality work; the general tier (Claude sonnet-4-6) handles high-frequency planning / validation / Vision-Audit; the -flash tier keeps draft-compare & basic polish at sub-second latency.
• MiniMax safety net: any primary LLM / video / image failure auto-routes to MiniMax (OpenAI-compatible) — surfaced live on the API Health Board (正常 / 额度用尽 / 配置缺失 / 不可达).
• Swap anything in .env.local (OPENAI_* / OPENAI_CREATIVE_* / VEO_* / IMAGE_MODEL / VE_TTS_MODEL / MINIMAX_*) — zero code change. See docs/llm-providers.md.

---

2. configure (3 mandatory lines, see docs/llm-providers.md for swaps)

cp .env.example .env.local

Edit .env.local:

Optional second terminal for real-time collab:

npm run dev:ws # Yjs WebSocket server on :1234

🩺 API Health Board — never get surprised by a dead key *(v6.7)*

Live status for every model and gateway: 正常 / 额度用尽 / 配置缺失 / 不可达, with real balance read-out and a "去充值 / 补配置" hint. Keys are never stored or returned. <p align="center"><img src="assets/v6/health.jpg" width="100%" /></p>

Zero-key trial (one command, **no API key needed**)

docker run -p 3100:3100 -e MOCK_ENGINES=1 -e JWT_SECRET=demo -e NEXTAUTH_SECRET=demo \
  ghcr.io/chrischen667788/wind-comic:latest

Then open <http://localhost:3100>. No clone, no npm install, no local build.

Prefer compose (also seeds a demo project):

docker compose -f docker-compose.demo.yml up

⚠️ Honest note: demo mode runs the deterministic mock engines (MOCK_ENGINES=1 — SVG storyboards, solid-colour clips, sine-wave audio). It demonstrates the pipeline and the product shape, and is not representative of real generation quality. Every artifact is labelled as mock — passing a placeholder off as a finished film is exactly the thing this project keeps hunting down. Add a single API key to switch to real engines.

OPENAI_API_KEY=sk-...

OPENAI_BASE_URL=https://api.openai.com/v1 # or any compat provider

📂 More modules — earlier (v6–v8) captures

Features that still ship. Director console · novel→season · finished-film station · team workspace · Cinema timeline are refreshed to v10 (live demo data). The style gallery, API health board, Polish Pro audit, and character turnaround are kept as earlier (v6–v8) captures because they show fuller sample output (the full style grid / live balances / a complete Pro audit / a 3-view turnaround sheet).

1. **Multi-agent pipeline, not one black-box model**

Director plans the story → Writer drafts dialogue under McKee structure → Style Bible Frame locks the look → Character Designer extracts an 8-dimension DNA signature of each character → Storyboard renders with Vision Audit (auto-regen on <70 score) → Video producer races multiple engines (Minimax / Veo / Kling) → Editor cuts j/l-cut on emotional beats and burns CJK subtitles.

🥊 vs. competitors

Lineup verified 2026-08-31 (Artificial Analysis arena; 6 parallel web-research lenses + key figures independently re-checked. Full analysis: docs/COMPETITIVE-GAP-2026-08.md): the top slot changed hands, and a Chinese model took it — T2V-with-audio: Wan 3.0 (1242) → Gemini Omni Flash (1237, down from 1244). New entrant xAI Grok Imagine Video 1.5 (Jun 16): 7 simultaneous visual reference anchors (character + scene + prop + style) plus native audio and voice cloning, $0.08 (480p) / $0.14 (720p) / $0.25/s (1080p) — more reference slots than our current cref+sref pair. Kling 3.0 remains the all-round pick for short drama (native 4K / 60fps / 15s / 6 coherent shots / built-in multilingual dialogue + lip-sync, $0.084–0.112/s); Veo 3.1 (first shipped 2025-10-15; 4K and vertical added 2026-01-13) can reach 60s+ via Scene Extension — but that is stitching; the single-generation ceiling is still 8–10s. It uses its own joint audio-visual generation (Lyria 3 is a separate Google music model, not Veo's audio component), from $0.40/s. Note: Veo 3.1's own T2V+audio Elo is ~1091 — the 1237 on the board belongs to Gemini Omni Flash, a different product. ✅ The Sora 2 API shutdown (2026-09-24) is confirmed by OpenAI — and this project has been guarded since v12.173/207: warn before the date, auto-drop from the model chain after it (falling back to veo/kling), throw only if the chain empties (tests/v12-173-sora-sunset.test.ts locks it). Not an open action item. 🇨🇳 The domestic camp broke through on both shot length and reference count — and two engines we already ship are behind. Seedance 2.5 (7-31) and Wan 3.0 (8-24 GA) both do native 30s single takes; Seedance accepts 50 reference inputs (token-billed, ¥42–70/M tokens), Wan 3.0 runs ¥0.42/s at 720P and converts documents straight to video (PPT/Word/PDF → 30s). Vidu Q3 is the only Chinese model explicitly positioned for short-drama industrialisation (7 reference images with multi-subject locking, 6 cinematic effect packs, lip-sync driving; Turbo 1080p ≈¥0.41/s). PixVerse C1 (4-08) is the most narrowly short-drama-vertical of all (storyboard-grid output + multi-speaker lip-sync). ⚠️ Two findings about our own stack, which matter more than any competitor: ① services/minimax.service.ts:240 defaults to MiniMax-Hailuo-2.3 (nothing overrides it in .env.local, so that is what actually runs) — but 2.3 / 2.3-Fast are marked legacy by the vendor; H3 (V2 API) is the recommended path, and this vendor has pulled an endpoint before with no notice (Music API: 410 + 2153). We list MiniMax H3 as a competitor column in the table below while still calling its previous generation. ② our two Vidu call paths disagree: qyt-vidu.service.ts pins viduq3, while vidu.service.ts sends no model field at all and rides whatever the vendor defaults to — the same disease as Midjourney「no version pinned anywhere, gateway default wins」: when the vendor changes its default our behaviour changes silently, and we cannot reconstruct afterwards which model produced a given shot. Both are queued as v12.402 / v12.403. <sub>This round re-checked every high-risk figure with an independent second search, and two claims I had already written into this README were overturned: Veo 3.1 first shipped 2025-10-15 (2026-01-13 was a feature update), its 60s+ is Scene Extension stitching — the single-generation ceiling is still 8–10s, and Lyria 3 is a separate Google music model, not Veo's audio component; and Runway Gen-4.5 has had native audio since 2025-12-11 — the earlier draft called its absence a fatal flaw, which was simply wrong. Elo figures are arena snapshots and move with voting.</sub> ⭐ BYO 架构再次接住这波:榜上模型基本都开放 API,填 key 即成为本管线可调度的引擎 —— 竞品越强,本管线越强。

🔴 We overclaimed last round — four "only we have this" claims retracted (verified 2026-09-03). This is not competitors catching up; the claims were too broad to begin with: ① "EDL/AAF export is unique to us" — Descript has had it: Timeline Export covers Premiere XML, FCPXML, Reaper (EDL) and Pro Tools/Logic (AAF). ② "the only MIT, self-hostable end-to-end AI short-drama platform" — false: EvoLinkAI/ai-short-drama (MIT + Docker + novel→script→storyboard→image→video→voice→film) qualifies too, alongside LocalMiniDrama, Toonflow, huobao-drama and others. ③ "BYO multi-provider registry is unique" — PopShort.AI already ships cross-vendor models (Veo 3.1 / Kling / Seedance / Nano Banana Pro). ④ "no competitor does pacing/reversal auditing" — Volcano Drama publicly claims a multi-agent verification mechanism with 200+ shot strategies optimising conflict, reversal and climax. What survives, stated narrowly: Volcano Drama optimises inline during generation, and Descript exports but has neither pacing audit nor an open licence — so as of 2026-09-03, no single product has all three of「pacing audit hardened into an independent blocking engineering gate + EDL/AAF export + open-source self-hostable for commercial use」. Note this is narrower than last round: "pacing audit" alone no longer holds; it must be "an independent blocking gate". Open-source scale: OpenMontage 54.7k★ (AGPL), ViMax 12.2k★ (MIT), DramaClaw 4.8k★ (Elastic 2.0, not FOSS), BigBanana 1.8k★ (non-commercial), Novella AI 89★ (MIT). Full analysis: docs/COMPETITIVE-GAP-2026-09.md. v12.214→244 双线推进。产品层:GPT Image / Nano Banana(Gemini)接入插件式图像 provider 链(issue #11,社区 @flobo3 提议,OPENAI_IMAGE_ENABLED / GEMINI_API_KEY 门控、原生 i2i 接角色一致性契约);多集连续生成补上「剧情记忆」(第 N 集 Writer 注入前几集前情提要 + 承接纪律,对标红果/阅文的 60~100 集连续,此前各集独立成篇)。平台/工程层:六轮独立对抗复检把安全洞从 CRITICAL 到 LOW 全清(SSRF 逐跳重验重定向 + IPv6 全隧道变体 / serve-file 签名能力 URL / WebSocket 鉴权 / 预算护栏),并把反复踩的「改了守卫却没跟到消费方」这个病固化成 CI 入库门禁(npm run gate:consumer,零容忍,上线即抓到 2 个人肉复检漏掉的真 SSRF)。 结论不变:生成层已是红海(竞品在出片/多镜/音频都第一梯队),Wind Comic 护城河收窄到「制作/平台层」——节奏审计、智能剪辑、字幕烧入、协作、自托管、开源、BYO。

Cells marked ⚠️ = the feature exists but in a limited / locked-down form (e.g. "you can only do this on a paid Pro tier through a UI panel").

---