Clone any viral video with AI agents. Not just a script, the whole workflow: swap the face, the words, the B-roll, ship 100 variants in one command, and get your 100M views.
TypeScript agentic-aiaiai-agentsai-videocompilerdslffmpeggenerative-aillmmarkup-languagemonorepoplugin-systemprogramming-languagetext-to-videotypescriptvideovideo-automationvideo-clonevideo-editingvideo-generation
docs: tighten the architecture overview
kashorinbde05e8fb62e
diff --git a/README.md b/README.mdindex ca076eca..60a0290b 100644--- a/README.md+++ b/README.md@@ -147,9 +147,9 @@ Ninety years later, Narratage gives the same idea a new form: a language and sys## Architecture-AI generation is slow, costly and non-deterministic. Narratage freezes requested work into a finite plan, then advances it from durable state. A build can pause, retry or change implementations without losing completed artifacts or repeating paid work.+AI generation is slow, costly and non-deterministic, so Narratage compiles requested work into a durable plan that can pause, retry and resume without repeating completed work.-At the center is a deliberately small, domain-neutral Core. It compiles requested outputs and explicit implementation choices into a plan, accepts execution events and decides what becomes ready next. It does not parse SVML, understand video, call providers, access credentials or render frames.+The domain-neutral Narratage Core only compiles plans and advances Build state. Parsing SVML, calling models and rendering frames all live in packages.Installing a package adds capability without requiring a Core release.@@ -163,8 +163,6 @@ Installing a package adds capability without requiring a Core release.| **Runtime** | Scheduling, storage, credentials and execution | `runtime`, `store-sqlite`, `local` || **Applications** | User-facing ways to author and operate Narratage | `cli`, `svml-playground` |-SVML imports define what the source means. The Runtime Profile separately selects packages that may access files, networks and credentials.-## Where to go next- [Demos](https://narratage.hypit.ai/) — see SVML source and its rendered result side by side.diff --git a/README.zh-CN.md b/README.zh-CN.mdindex 97bd6fcb..a0b5dfa9 100644--- a/README.zh-CN.md+++ b/README.zh-CN.md@@ -139,9 +139,9 @@ Narratage 为文件规定不同的角色,但不保留任何固定文件名:## 架构-AI 生成缓慢、昂贵且不确定。Narratage 把被要求的工作冻结为一份有限计划,再依据持久化状态推进。Build 可以暂停、重试或更换实现,而不会丢失已完成的产物,也不会重复执行已经付费的工作。+AI 生成缓慢、昂贵且不确定,因此 Narratage 把所需工作编译为一份持久化计划,可以暂停、重试和继续,而不会重复已经完成的工作。-系统中心是一个刻意保持极小、领域无关的 Core。它把被要求的输出和显式选择的实现编译为计划,接受执行事件,并决定下一步哪些工作可以运行。它不解析 SVML,不理解视频,不调用 Provider,不访问凭据,也不渲染画面。+领域无关的 Narratage Core 只负责编译计划和推进 Build 状态。解析 SVML、调用模型和渲染画面全部存在于包中。安装一个包即可增加能力,无需重新发布 Core。@@ -155,8 +155,6 @@ AI 生成缓慢、昂贵且不确定。Narratage 把被要求的工作冻结为| **Runtime** | 调度、存储、凭据与执行 | `runtime`、`store-sqlite`、`local` || **Applications** | 创作和操作 Narratage 的用户界面 | `cli`、`svml-playground` |-SVML 的导入决定 Source 表达什么;Runtime Profile 另外选择可以访问文件、网络和凭据的包。-## 接下来去哪- [演示](https://narratage.hypit.ai/zh/) —— 并排查看 SVML 源码与它渲染出的画面。