baocut README

项目的 README,来自 github.com/JimLiu/baocut。合并为一个 markdown 文件:README.md。

BaoCut

BaoCut

The video editor built for agents.
Your agent does the editing. You review and fine-tune.

English · 简体中文


Hand BaoCut to your agent (Claude Code, Codex, Cursor and others) and say what you want in one sentence: transcribe, subtitle, translate and dub, cut a talking-head recording, or make an explainer from scratch. The agent edits through commands, and the result is a timeline you can keep editing. Open the editor, watch it once, then tell the agent what to change or drag things yourself.

Quick start

1. Install BaoCut and open it once

Download for macOS (Apple Silicon) or Windows. On first launch the app puts the baocut command in your terminal. Check it:

baocut status

It lists what this machine can do right now and what is missing (local models, ffmpeg and so on), each with a fix.

If the terminal cannot find baocut (on Windows, open a new terminal first), install it once by full path (macOS shown); it tells you whether a folder still needs adding to PATH. For other cases see the desktop CLI entrypoint.

/Applications/BaoCut.app/Contents/MacOS/baocut-cli cli install

2. Install the BaoCut skill for your agent

baocut skill install

The skill is a manual on how to use BaoCut. Once it is installed, your agent knows when and how to use it. It goes into ~/.agents/skills and ~/.claude/skills by default and updates with the app; use --dir <folder> for another location. For agents that connect over MCP, run baocut mcp install.

3. Ask your agent

Transcribe ~/Movies/talk.mp4, add English and Chinese subtitles, and export an MP4.
Download the video at this link, translate it into English and dub it: <video page link>
Cut the filler words and long pauses from this recording, then make 3 vertical shorts.
Use BaoCut to make a 3-minute explainer about <topic>, presented by a virtual host.

4. Review and fine-tune

When the timeline plays from start to finish, the agent stops and gives you a review link. It opens this video's editor in your browser. Say what to change ("the subtitle at 0:42 covers the face"), or drag clips and fix captions yourself; you and the agent edit the same timeline. Once you approve, it exports the final video.

No external agent? That works too. Chat in the desktop app's Home. It runs on the Codex CLI or Claude Code signed in on this machine, and works the same way.

CleanShot 2026-10-07 at 9 35 49 PM@2x CleanShot 2026-10-07 at 9 38 02 PM@2x

Why "built for agents"

Traditional editors are designed for human hands, so an agent has to work around the interface. Most AI video tools go the other way: they render one MP4, and if the second sentence of the translation is wrong, you start over. BaoCut lets the agent edit through commands and keeps the result on a timeline:

What it can do

From existing footage

From a brief

Delivery

Four ways in, one Runtime

Every surface talks to the same local Runtime over the same protocol and edits the same video.

Surface What it is
Desktop app (Electron) Home for working with the agent, Space for browsing and re-editing everything a project produced, a full timeline editor with preview, undo and live captions.
CLI baocut Thin client for scripts and terminals. Emits JSON when piped; exit codes tell you whether the user needs to act.
MCP service + BaoCut skill Your own agent (Claude Code, Codex, Cursor, Gemini CLI…) drives BaoCut through the same tool catalog. baocut skill install links the skill into ~/.agents/skills and ~/.claude/skills (or any folder) so it updates with the app; baocut mcp install wires up MCP. Same tool names on both paths.
Web client Open any video's editor in the browser with a one-time link from baocut web open --video <id> so a remote agent can show you its work.

Bring your own agent engine. Claude Code and Codex CLI are validated end to end. GitHub Copilot CLI, Pi and OpenCode are built in, and any Agent Client Protocol agent (Qwen Code, Gemini CLI, Cursor Agent, Grok, Kimi Code, or one you add) plugs in through the same driver.

Bring your own models. Capabilities (transcribe, speak, generate image, generate text, separate audio) are decoupled from providers. Run local models on Apple Silicon (MLX, Core ML) or on Windows (whisper.cpp + candle, with CUDA and Vulkan builds), with local speech (Qwen3-TTS, IndexTTS, VoxCPM2, OmniVoice, Fish Audio S2 Pro; Built with Fish Audio, research/non-commercial license) and music on the bundled audio.cpp engine, or point a capability at OpenAI, Anthropic, Google, ElevenLabs, DeepSeek, Moonshot, Qwen, Zhipu, MiniMax, Volcengine, xAI, Mistral, Groq, OpenRouter, SiliconFlow or any OpenAI-compatible endpoint. A usage ledger tracks cost per job. One machine can share its local models with others on the LAN.

Download and install

BaoCut is available for macOS and Windows. The links below always download the latest release; version numbers and release notes are on baocut.com and GitHub Releases. Packaged apps include the Runtime and native workers; Node.js and Rust are only needed for development.

Platform Installer ZIP Choose this version for
macOS 14+ · Apple Silicon DMG ZIP Apple Silicon Macs; Developer ID signed and notarized
Windows x64 · CPU Setup EXE ZIP CPU inference
Windows x64 · CUDA Setup EXE ZIP NVIDIA GPUs supported by CUDA 13, starting with RTX 30 / Ampere
Windows x64 · Vulkan Setup EXE ZIP Vulkan-capable AMD / Intel GPUs or older NVIDIA GPUs; Whisper and audio.cpp (speech, music) use the GPU, candle uses the CPU

On macOS, open the DMG and drag BaoCut into Applications. On Windows, install the Microsoft Visual C++ v14 x64 Redistributable first, then run the installer or extract the ZIP. Windows packages are unsigned; GPU variants need compatible graphics drivers. Checksums and verification details are on each release's GitHub page. The app checks GitHub Releases for updates through its platform-specific update feed.

External agents use the desktop CLI entrypoint; see that guide for absolute paths, BAOCUT_CLI, and older packages that lack it.

Agent features require a signed-in agent engine such as Codex CLI or Claude Code. Media analysis, transcription preparation and export also require ffmpeg and ffprobe; these external tools are not bundled.

Local models download from Hugging Face by default. If it is unreachable or slow, set the download source to ModelScope (Download and storage at the bottom of a model page's local models, or Settings › General › Model download source) to download from ModelScope instead (the few model files ModelScope lacks still come from Hugging Face; every file is checked against its built-in checksum), or choose Custom and enter a mirror's base URL.

Run from source

For normal use, choose a packaged app: it includes the Rust workers and WASM, so you do not need Node.js, Rust or a compiler. Source development has two paths:

Goal Command after npm ci Requirements / limits
Full desktop development npm run dev Node.js 24.22.1+ (LTS), Rust, CMake and platform build tools below. Builds native workers and WASM before launching.
Shell/UI development without Rust npm run dev:lite Node.js 24.22.1+ (LTS). Skips native and WASM builds. On a fresh checkout, video editing, preview, export, local inference and speech processing are unavailable; existing outputs may still be used.

Agent conversations additionally need an installed, signed-in agent engine. Media analysis, transcription preparation and export need ffmpeg and ffprobe on PATH. Neither is bundled. Opening the app does not require an agent login.

1. Install the tools for your OS

Install Git and Node.js 24.22.1+ (LTS) first. Open a new terminal and check git --version, node --version and npm --version. For lite mode, skip the Rust/compiler steps below and continue at step 2.

macOS

Install the Command Line Tools (xcode-select --install). If you use Homebrew, install CMake and FFmpeg with brew install cmake ffmpeg. Install Rust using the official rustup installer:

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
. "$HOME/.cargo/env"
rustup default stable
rustup target add wasm32-unknown-unknown

On Apple Silicon with macOS 14+, the local model worker uses MLX / Core ML and also needs full Xcode with the Metal compiler. Install Xcode, launch it once to finish setup, and select it (adjust the path if installed elsewhere):

sudo xcode-select --switch /Applications/Xcode.app/Contents/Developer
xcrun --find metal
# If Metal is missing from the selected Xcode:
xcodebuild -downloadComponent MetalToolchain

Intel Macs use candle / whisper.cpp instead, requiring CMake and a C++ compiler. Published macOS packages target Apple Silicon; Intel source builds are not covered by the release checks.

Windows x64

Install Visual Studio / Build Tools with Desktop development with C++, the MSVC x64 tools and a Windows SDK (see Rust's MSVC prerequisites). Install CMake with its PATH option enabled, and LLVM for the libclang used by Rust bindings. If bindings cannot locate it, set LIBCLANG_PATH to LLVM's bin directory. Install FFmpeg with both ffmpeg.exe and ffprobe.exe on PATH.

Download and run rustup-init.exe, keeping the default x86_64-pc-windows-msvc host. Reopen an x64 Native Tools Command Prompt or a Developer PowerShell configured for x64 before running the source-build commands; a regular terminal may find Rust but lack cl.exe and the SDK environment.

rustup default stable
rustup target add wasm32-unknown-unknown
cargo --version
cmake --version
cl /?

The default Windows build uses CPU inference and does not need CUDA or Vulkan SDKs. GPU builds have extra requirements; see the desktop guide.

Linux (source development)

Install Rust with the same rustup commands as macOS. On Debian/Ubuntu, install the native tools with sudo apt install build-essential cmake clang libclang-dev pkg-config ffmpeg. Other distributions use their corresponding packages. Linux source development uses candle / whisper.cpp; Linux installers and native release validation are not provided.

2. Clone, check and run

Run from the repository root. These commands work in macOS/Linux terminals and Windows developer terminals:

git clone https://github.com/jimliu/baocut.git
cd baocut
npm ci
npm run doctor
npm run dev

Use Node.js 24.22.1 LTS (pinned in .node-version and CI). Patched Node.js 22.23.4+ and 26.12.1+ are also supported within their respective major versions; older releases are rejected by doctor because they lack the October 2026 security fixes. The npm scripts that run TypeScript source explicitly enable type stripping. doctor reports missing tools with installation hints and exits nonzero for missing source-build prerequisites. FFmpeg is a warning because it is needed for media workflows, not to compile the app. The check does not install tools or prove all SDKs and Rust dependencies can build. npm run setup checks prerequisites and builds native workers and WASM without opening Electron; npm run dev runs it automatically and stops if either build fails. Rustup-based WASM builds install the target when missing. The first build downloads dependencies and can take considerable time and disk space; later runs reuse Cargo's cache. Cargo is also discovered in CARGO_HOME/bin and next to rustup when absent from PATH.

Without Rust, use this instead of doctor and dev:

npm run dev:lite

After installing the full toolchain, run npm run dev to restore the native features. Development mode starts Electron and Vite, launches the Runtime on demand and stops it on exit. Data goes to .dev/baocut-home in the repo. Electron 44 downloads its binary before dev, dev:lite or start, so first launch needs network access. The development supervisor asks Electron, Vite and the Runtime to exit on Ctrl+C or terminal closure, then force-kills remaining processes after 12 seconds.

Troubleshooting

To connect a CLI to the Runtime started by the development desktop, on macOS/Linux:

BAOCUT_HOME=.dev/baocut-home npm run cli -- status

On Windows PowerShell:

$env:BAOCUT_HOME = '.dev/baocut-home'
npm run cli -- status

In Windows Command Prompt, use set "BAOCUT_HOME=.dev/baocut-home" before the CLI command.

With the development desktop closed, the same command starts the repository's Runtime in that home, so tools work before they are released. Without BAOCUT_HOME, the CLI uses your real ~/.baocut and the installed app's Runtime, and commands newer than that app report RUNTIME_OUTDATED.

For a separate home to test in, for example with an agent, run npm run cli -- sandbox create ~/baocut-sandbox (add --link-accounts-from ~/.baocut to link your model accounts read-only; calls are billed to them). It creates the folder, a bin/baocut that runs this checkout's CLI on it, and env.sh / env.ps1. After . ~/baocut-sandbox/env.sh, baocut in that shell and agents started from it use the sandbox.

Other commands

Command What it does
npm run doctor / npm run setup Check source-build prerequisites / check and build native workers, fetch the bundled audio.cpp engine, and build WASM
npm run dev:lite Run the desktop shell without Rust builds; see the limits above
npm run dev:designs Start the interactive prototype with Vite at http://127.0.0.1:4331/#/home, rebuilding and reloading on source changes. First run npm --prefix designs/baocut ci; see the prototype README.
npm run build Build main process, preload, Runtime and UI into apps/desktop/out
npm start Preview the already-built desktop app (UI loaded from files); first run npm run setup and npm run build
npm run package:mac -- --build <n> --sign-sha1 <SHA1> --out <new-directory> Signed and notarized Apple Silicon ZIP and DMG; see the desktop guide.
npm run package:win / package:win:cuda / package:win:vulkan Windows x64 installer (NSIS, per-user) and zip, unsigned: CPU, CUDA (NVIDIA) or Vulkan (AMD / Intel) model worker. See desktop README
npm run runtime Start the Runtime alone (default home ~/.baocut)
npm run cli -- --help The baocut CLI. Tool commands derive from the Runtime's tool catalog (same as MCP); baocut help <command> for arguments, baocut spec for a machine-readable catalog. Starts a Runtime in the background when none is running and lets it exit when idle (baocut runtime status|ensure|stop). Exit codes: 0 ok, 1 failed, 2 user action needed, 3 Runtime unavailable, 4 bad arguments. Design in Agent surface §5
npm run build:engine Build the video engine (engine-host), export renderer (export-worker), speech-worker, face detection (face-detect), background removal (matte) and local inference process (model-worker). Backend follows the platform: MLX and Core ML on Apple Silicon, candle and whisper.cpp elsewhere (needs CMake and a C++ compiler); LiteRT-LM text embedding everywhere. Extra args go to cargo (-- --release)
npm run build:audiocpp Download the pinned audio.cpp release (local speech and music engine) for this machine into the cargo output dir's audiocpp/, checking size and sha256; -- --variant cpu takes the CPU build on Windows / Linux. BAOCUT_TOOLS_ENDPOINT downloads from a mirror. Packaging bundles the same release
npm run build:litertlm-runtime On Windows x64, unpack the LiteRT-LM C library that BaoCut builds for local semantic search (packages/models/vendor/litertlm-runtime/) into the cargo output dir's litertlm-runtime/, checking size and sha256; does nothing on other platforms (macOS downloads it with the model). Packaging bundles the same file
npm test Unit tests and Runtime end-to-end tests with a fake driver (no Codex needed)
npm run check:file-fonts Build the desktop app and verify, in a hidden Electron window over file://, that the preview renderer loads every font
npm run typecheck Type check
npm run build:slides-editor Build the slides editor (packages/slides-runtime) into apps/slides-editor/dist: the editor page and the compiler slides_check uses. Skipped when inputs are unchanged (node apps/slides-editor/build.mjs --force rebuilds). See slides design §3
npm run build:design-editor Build the design (canvas) editor (packages/design-runtime) into apps/design-editor/dist: the editor page, the artboard runtime and the fast-edit worker. Skipped when inputs are unchanged (node apps/design-editor/build.mjs --force rebuilds). See packages/design-runtime
npm run build:design-system-editor Build the design system editor (packages/design-system-runtime) into apps/design-system-editor/dist: the editor page, the preview page artifact-type/run.html and the React UMD files previews load. Skipped when inputs are unchanged (node apps/design-system-editor/build.mjs --force rebuilds). See packages/design-system-runtime
npm run build:docs-editor Build the docs editor (packages/docs-runtime) into apps/docs-editor/dist: the editor page, its runtime and fonts. Skipped when inputs are unchanged (node apps/docs-editor/build.mjs --force rebuilds). See docs design §3

Status

BaoCut is under active development, with macOS Apple Silicon and Windows x64 packages available above. What is true today:

Learn more

docs/ Product design, architecture, format and protocol specs, acceptance criteria. Start with the reading guide.
designs/baocut/ Interactive UI prototype; UI changes land here first.
skills/ The BaoCut skill: the entry that teaches agents how to use BaoCut over CLI or MCP, with the built-in craft skills (subtitles, translation, talking-head cut, narration, motion graphics…) under skills/baocut/built-in-skills/.
templates/ Built-in creation templates.
AGENTS.md Rules for coding agents working in this repository.

Environment variables
Variable Purpose
BAOCUT_LEGACY_ROOT Explicit v1/v2 data directory for startup migration, including in an isolated development home. Old data stays unchanged; completion is recorded in <home>/store/legacy-upgrade.json. The default desktop launch (development or packaged) detects historical platform directories automatically; explicitly configured sandbox homes stay isolated
BAOCUT_HOME Runtime home: discovery file, instance lock, session store, logs. Default ~/.baocut; .dev/baocut-home in desktop development mode
BAOCUT_PROJECTS_DIR Where "New project" creates directories. Default ~/BaoCut; <home>/projects when BAOCUT_HOME is set
BAOCUT_<ID>_PATH Executable of an agent engine; <ID> is one of CLAUDE, CODEX, GEMINI, CURSOR, GROK, KIMI (e.g. BAOCUT_CODEX_PATH). Otherwise the newest version found on the login shell's PATH and known install locations
BAOCUT_LOCALE Language of UI, Runtime and CLI text (en, zh-Hans; zh-CN and en_US.UTF-8 are accepted). Overrides the setting. For tests and troubleshooting; the test suite sets zh-Hans
BAOCUT_SYSTEM_LANGUAGES System preferred languages as the Runtime sees them, comma separated. The desktop app passes the OS setting; otherwise LC_ALL / LC_MESSAGES / LANG
BAOCUT_ALLOWED_ORIGINS Extra origins allowed to connect to the Runtime, comma separated. Desktop development mode adds Vite's address
BAOCUT_BIN_DIR Directory of the native programs shipped with the app (workers and credential helper). When set, cargo outputs are not searched; the packaged desktop app sets <resources>/bin
BAOCUT_ENGINE_HOST / BAOCUT_MODEL_WORKER Path to engine-host / model-worker. Otherwise BAOCUT_BIN_DIR, then cargo output dirs (CARGO_TARGET_DIR, .cargo/config*.toml build.target-dir, repo target/) under {release,debug}/. Without model-worker, local transcription is unavailable
BAOCUT_EXPORT_WORKER Path to export-worker. Otherwise next to engine-host, then the cargo output dirs; without it export fails with EXPORT_TOOL_MISSING
BAOCUT_FACE_DETECT Path to face-detect. Otherwise BAOCUT_BIN_DIR, then the cargo output dirs; without it (or without the shipped YuNet model) videos_frames still returns frames, with detector: null and no faces
BAOCUT_MATTE Path to matte (background removal). Otherwise BAOCUT_BIN_DIR, then the cargo output dirs; without it the assets_process tool's remove-background is refused
BAOCUT_RUNTIME_ENTRY Entry the CLI uses to start a Runtime: .ts / .js run with the current Node, anything else as an executable. Otherwise the app the CLI ships in; apps/runtime when running from the repo with a non-default BAOCUT_HOME; an installed BaoCut app; then apps/runtime in the repo
BAOCUT_MODELS_DIR Local model files. Default <BAOCUT_HOME>/models
BAOCUT_MODELS_ENDPOINT Download source (base URL) for local models; overrides the Model download source setting. Default https://huggingface.co; https://www.modelscope.cn downloads from ModelScope
BAOCUT_TEMPLATES_DIR Built-in creation templates (template spec §6). Otherwise the packaged templates/, then the repo's. User templates live in <BAOCUT_HOME>/templates; built-in wins on id clash
BAOCUT_SKILLS_DIR Built-in agent skills (architecture §3.8). Otherwise the packaged skills/baocut/built-in-skills/, then the repo's. User skills live in <BAOCUT_HOME>/skills; built-in wins on id clash
BAOCUT_AGENT_SKILLS_DIR Source of the BaoCut skill for external agents (Agent surface §8), rendered by the Runtime. Otherwise the packaged skills/, then the repo's
BAOCUT_AGENT_SKILL_DIR The rendered BaoCut skill that baocut skill install links to; the desktop app sets it to its agent-skill/baocut. Without it (development) the Runtime renders into <BAOCUT_HOME>/agent-skills/baocut
BAOCUT_CLI_ENTRY The app's own CLI launcher, shown by baocut skill status and the Skills page; the desktop app sets it. Unset in development
BAOCUT_SLIDES_EDITOR_DIST Built slides editor (slides design §3). Otherwise the packaged slides-editor/, then the repo's apps/slides-editor/dist
BAOCUT_DESIGN_EDITOR_DIST Built design editor (packages/design-runtime). Otherwise the packaged design-editor/, then the repo's apps/design-editor/dist
BAOCUT_DESIGN_SYSTEM_EDITOR_DIST Built design system editor (packages/design-system-runtime). Otherwise the packaged design-system-editor/, then the repo's apps/design-system-editor/dist
BAOCUT_DOCS_EDITOR_DIST Built docs editor (docs design §3). Otherwise the packaged docs-editor/, then the repo's apps/docs-editor/dist
BAOCUT_MODEL_ASSETS_DIR Model data shipped with the app (self-test samples, reference recordings of built-in voices, i.e. packages/models/assets). Otherwise the packaged model-assets/, then the repo's; the packaged desktop app sets <resources>/model-assets
BAOCUT_WORKER_FEATURES Extra cargo features for model-worker in npm run build:engine, comma separated. cuda,whisper-ggml-cuda for NVIDIA on Windows / Linux (needs the CUDA toolkit; Whisper on CUDA untested), whisper-ggml-vulkan for AMD / Intel (needs the Vulkan SDK; untested)
BAOCUT_GPU off, 0, false, cpu or no keeps model-worker on the CPU for both candle and whisper.cpp (Model Worker protocol §2.1)
BAOCUT_AUDIOCPP_CLI Path to an audiocpp_cli you built; audiocpp_server is taken from the same directory. When set, only it is used (a missing file means no engine). Otherwise the bundled <resources>/bin/audiocpp/, in development the cargo output dir's audiocpp/ (Model Worker protocol §4.4)
BAOCUT_AUDIOCPP_BACKEND audio.cpp backend for local speech and music (metal, cuda, vulkan, cpu). Otherwise the CLI picks its own and the speech server uses the first of Metal, CUDA, Vulkan compiled into the build, else CPU
BAOCUT_FFMPEG ffmpeg used for media analysis, WebM playback cache and transcription audio preparation. Otherwise found on the login shell's PATH
Repository layout
apps/
  desktop/          Electron main process, preload, UI entry
  runtime/          Runtime process entry
  cli/              Command-line client
  web/              Web client, served by the Runtime's web service
  slides-editor/    Slides editor page (built from packages/slides-runtime), served by the Runtime
  design-editor/    Design (canvas) editor page (built from packages/design-runtime), served by the Runtime
  design-system-editor/ Design system editor page (built from packages/design-system-runtime), served by the Runtime
  docs-editor/      Docs editor page (built from packages/docs-runtime), served by the Runtime
  portal/           baocut.com website: static pages and a Cloudflare Worker for downloads
packages/
  protocol/         Wire protocol: envelopes, methods, topic events, domain types, validation
  client/           Connect, request, subscribe, reconnect; mirror reduction
  runtime-core/     Assembly, WebSocket gateway, request handling
  harness/          Sessions, tasks, stop barrier, Driver ABI
  agent-drivers/    Agent engine adapters: Codex app-server, Claude Agent SDK, Copilot, Pi, OpenCode, ACP
  runtime-storage/  Runtime home, session and project storage, discovery file
  models/           Capabilities, local model bundles, selection, usage ledger
  providers/        Online providers: OpenAI, Anthropic, Google, ElevenLabs, OpenAI-compatible
  jobs/             Long-running jobs and scheduling
  nodes/            LAN capability sharing (node service and initiator)
  code-runtime/     Code bundle adapters and frame baking
  ui/               React UI (React Spectrum 2 + Zustand)
  slides/           Slide decks: file protocol, deck store, editor routes, editor host shim
  slides-runtime/   Slides editor runtime: document model, HTML compiler, editor UI, presentation, export
  design/           Design canvases: file layout, design_check, editor host shim
  design-runtime/   Design editor runtime: canvas and artboards, properties, preview documents, artboard runtime, export
  design-system/    Design systems: directory layout, design_system_check, editor host shim
  design-system-runtime/ Design system editor runtime: tokens, components, assets, docs, previews, files-layout storage
  docs/             Docs: per-directory doc store, prose model and ops, door and mux service, agent connector, editor host shim
  docs-runtime/     Docs editor runtime: tabs, rich text, tables, comments, dropdowns, export
crates/
  editor-semantics/ Time semantics: exact rationals, decimal seconds, frame quantization
  video-model/      Video interchange DTOs
  video-engine/     Video engine and storage: transactions, receipts, undo
  timeline/         Timeline and element model
  render-graph/     Frame plans shared by native export and WASM preview
  engine-host/      Engine Host process, talks to the Runtime over stdio
  export-worker/    Final render
  model-worker/     Local inference (ASR, image, separation, embeddings)
  speech-doc/       Transcript post-processing, alignment, subtitle splitting
bindings/
  preview-wasm/     WASM entry for the editor preview
skills/             The BaoCut skill (entry) for agents
  baocut/built-in-skills/  Built-in craft skills, indexed in its README
templates/          Built-in creation templates
tools/              Build scripts

Dependency direction and layer responsibilities: architecture §11, §13. Placement and naming: repo conventions.

License

BaoCut is source-available under the BaoCut Community License 1.0. In short:

baocut (JimLiu/baocut)

正在搜索数据库…

1.8K ·GitHub
这是什么?BaoCut 是一款 AI 视频代理,核心是可编辑的时间线,而不是一次性生成的 MP4。你告诉它想要什么,它就能完成转录、加字幕、翻译、配音、剪辑和动画制作,所有结果都会回到时间线中,方便你继续修改。
这是桌面应用,还是要通过编程代理与它交互?它是一款原生桌面应用(基于 Electron),带有完整的时间线编辑器。此外还提供 CLI、带有 BaoCut 技能的 MCP 服务,以及网页客户端。这四种方式都通过相同的协议连接同一个本地 Runtime,因此你可以自己编辑,也可以让 Claude Code 或 Codex CLI 等代理来完成。
需要安装 ffmpeg 或编程代理吗?需要。代理功能要求已登录的代理引擎,例如 Codex CLI 或 Claude Code。另外,媒体分析、转录和导出需要系统中有 ffmpeg 和 ffprobe,它们并未随应用一同打包。
能在 Linux 上运行吗?没有打包好的应用。Linux 仅支持通过源码进行开发,使用 candle/whisper.cpp,不提供 Linux 安装程序,也未进行发布验证。
BaoCut 和其他 AI 视频生成工具有什么不同?BaoCut 不是一次性渲染出最终文件交给你,而是将每一份转录稿、翻译、字幕、配音和动态图形都保留在同一个可编辑的视频中。修改某一行时,只需重新渲染对应部分;导出的文件也从来不是源文件。
没有现成素材,能根据文字提示生成视频吗?可以。它内置 24 种场景模板,包括讲解视频、产品发布、预告片、教程和数据故事。它还能从零开始规划脚本、合成旁白,并让画面与旁白对齐剪辑。
什么是 .baocut 文件包?它是 BaoCut 的便携项目文件,包含一个视频快照以及所有文档和素材版本,因此项目可以在另一台电脑上重新打开并继续编辑(不包括完整的事务/撤销历史记录)。
哪些编程代理可以驱动 BaoCut?Claude Code 和 Codex CLI 已完成端到端验证;GitHub Copilot CLI、Pi 和 OpenCode 已内置支持。任何兼容 Agent Client Protocol 的代理,例如 Qwen Code、Gemini CLI、Cursor Agent、Grok 或 Kimi Code,都可以通过同一驱动接入。
作者
JimLiu
大小
304.6 MB
星标
1.8K
时长
4个月
提交 · 90 天