Teach your agent
to make video
One sentence installs it into the AI assistant you already use. Research, script, voiceover, final cut: all on your own machine. Skills are free, editable, and shareable.
codex-macos-arm64.zh-CN.txt请帮我安装或更新 ScaleMuse。这个任务必须在 Codex Desktop 的 macOS 本地任务中执行；如果当前不是本地 macOS Apple Silicon 环境，请停止并告诉我，不能改用云端任务安装。 先读取 https://download.scalemuse.ai/runtime/latest.json，确认其中的平台是 macOS arm64、artifacts.dmg.latest_url 完全等于 https://download.scalemuse.ai/runtime/latest-macos-arm64.dmg、所有下载地址都属于 https://download.scalemuse.ai、签名状态是 ad_hoc 或 developer_id_notarized。然后把 artifacts.dmg.latest_url 指定的 DMG 下载到新建的临时目录，计算 SHA-256，并与 latest.json 中的 artifacts.dmg.sha256 完整比对。校验不一致时立即停止，不要打开或安装文件。 校验通过后，用 codesign --verify --deep --strict 验证 DMG 内“ScaleMuse Runtime.app”的签名完整性。若 latest.json 的签名状态是 developer_id_notarized，再额外验证 Apple 公证状态和 Gatekeeper 评估。若是 ad_hoc，请如实告诉我：这个版本由 ScaleMuse 自签名、未经 Apple 公证，完整性由上一步的 SHA-256 比对保证；ad_hoc 版本通不过 spctl 的静态评估是预期行为，不代表文件被篡改。不要移除 quarantine 属性，不要使用任何绕过 Gatekeeper 或关闭系统安全检查的方法。 在写入“/Applications”或覆盖已有 ScaleMuse Runtime 之前，请一次性告诉我：准备安装的版本、签名与校验结果、目标路径、是否会覆盖旧版本，以及需要的 macOS 权限，然后等待我确认。没有确认不要修改“/Applications”。 确认后，将“ScaleMuse Runtime.app”安装到“/Applications”，再运行： "/Applications/ScaleMuse Runtime.app/Contents/MacOS/scalemuse" setup codex 这一步会一次性安装或更新 ScaleMuse 官方 Skills，并创建稳定的 scalemuse 命令入口。完成后运行 Runtime 状态检查，报告 Runtime 版本、已安装或更新的 Skills，以及任何不可用能力。最后安全卸载 DMG 并删除本次创建的临时下载目录。
Open a local task in Codex Desktop and paste this as-is. Codex downloads the build, verifies the signature, asks once before writing to Applications, then installs every official Skill in one go. The notarized build is still being published; until then the agent stops at the first check and tells you, so nothing wrong gets installed.
- Skills and the local app are free
- Renders on your machine
- Credits only for voice and generation
Three steps from a sentence to a finished video
Copy one sentence, paste it into your agent
Your agent downloads the build, verifies the signature, and installs every official Skill in one go.
Say what video you want
"Make me a 45-second, 9:16 explainer about AI chips." The entry Skill picks the right method and checks in with you all the way to the final cut.
For exampleMake me a 45-second, 9:16 explainer about AI chips.The video renders on your computer
Project, timeline, and render all stay local. Credits are charged only when voice, image, or video models are used, and you see the cost before it happens.
Made with the same capabilities
These clips come from ScaleMuse voice, generation, and erase capabilities: the same ones every Skill renders through.
6 official Skills, one entry point
Just say "make me a video". The entry Skill hands the job to the right specialist Skill; the one marked "Tool" turns your own workflow into a new Skill.
Fork it, or write your own
A Skill is just a folder with a SKILL.md inside. Tell your agent one sentence and a workflow you built locally becomes a Skill anyone can install with one line.
Say one sentence to your agent, it does the rest
Scaffolding, copying an official Skill out to adapt, checking it against your local Runtime, publishing to the community catalog: this official Skill walks the agent through all of it and asks for your creator name on the first publish. It shows up on the Skills page the moment it is published.
Use $skill-author to turn my workflow into a Skill and publish it to ScaleMusePrefer typing commands yourself? Three steps:
Fork an official one
Copy it out and edit freely. Runtime upgrades never overwrite your changes.
scalemuse skill fork finance-tech-video my-financeStart from scratch
Scaffold a new Skill and let your agent write the method into SKILL.md.
scalemuse skill init my-skillShare it
Once published it shows up on the Skills page, and anyone can install it with one sentence.
scalemuse skill publish ./my-skillOthers install it with: scalemuse skill install @your-name/my-skill
Most of it is free. The paid part is spelled out.
- Skills and the local app: free
- Local projects, timelines, and rendering: free
- Voiceover, speech recognition, image and video generation: credits, quoted before you submit
Frequently asked questions
Do I need an account?
Not to install the Skills or the local app. You only need an API key, created in your profile, when your agent uses model capabilities like voiceover, image, or video generation.
Which agents and systems are supported?
Today: Codex, Claude Code, and Cursor on macOS (Apple Silicon). Skills install to ~/.agents/skills; Cursor reads them directly and Claude Code reads the same copy through a link. Windows and Linux are on the roadmap.
Do my files get uploaded?
No. Projects, timelines, and renders stay on your computer. Only the model capabilities you approve send the necessary inputs to ScaleMuse, and you see the estimated credits first.
Can I still make videos in the browser?
Yes. The web workspace (one-line creation, video remake, multilingual translation, smart erase) works as before and shares one account and credit pool with the Skills.