minimax-h3
calesthio/OpenMontage · Agent Skill
使用各種 API 和本地模型生成 4-15 秒的 2K MiniMax H3 (Hailuo 3.0) 影片。
前往安裝 ↓總覽
| 發布者 | |
|---|---|
| 儲存庫 | calesthio/OpenMontage ↗ |
| 技能資料夾 | minimax-h3 |
| 星 | 56,383 |
| 附帶檔案 | 1 |
| 授權 | AGPL-3.0 |
| 最近更新 | 2026年8月12日 |
| 連結 | 技能原始碼 ↗ · 儲存庫 ↗ |
Mozgi 檢查
我們在技能檔案中發現的內容。詳細資訊需訂閱。
安裝
技能就是一個含 SKILL.md 的資料夾。選擇你的代理,在專案根目錄的終端機(macOS、Linux 或 Windows 的 Git Bash)執行指令。
把提示貼到 Claude、Codex 或其他助理:它會自行檢查並安裝技能。
Install the “minimax-h3” skill from mozgi.io (https://mozgi.io/skills/minimax-h3/) for me. 1. Download the archive: https://mozgi.io/skills/minimax-h3/download If it fails, take the skill folder from the source: https://github.com/calesthio/OpenMontage/tree/main/.agents/skills/minimax-h3 2. Read SKILL.md and the other files. Check the skill is safe: no hidden commands, no sending data elsewhere, no scripts doing something other than described. If anything looks suspicious, stop and tell me. 3. Put the “minimax-h3” folder into your agent's skills folder: .claude/skills/ for Claude Code, .agents/skills/ for Codex, .cursor/skills/ for Cursor, .gemini/skills/ for Gemini CLI. Inside the project if we are in one, otherwise in the home folder (~/.claude/skills/ etc.). 4. Make sure <skills folder>/minimax-h3/SKILL.md exists. 5. Briefly tell me what this skill does and offer to try it on my task. If you cannot do a step yourself, tell me what to click or type.
git clone --depth 1 --branch main https://github.com/calesthio/OpenMontage.git /tmp/OpenMontage mkdir -p .claude/skills cp -r /tmp/OpenMontage/.agents/skills/minimax-h3 .claude/skills/minimax-h3
重新啟動 Claude Code — 技能就會出現在清單中,手動呼叫:/minimax-h3。要套用到所有專案,請放在 ~/.claude/skills。
git clone --depth 1 --branch main https://github.com/calesthio/OpenMontage.git /tmp/OpenMontage mkdir -p .agents/skills cp -r /tmp/OpenMontage/.agents/skills/minimax-h3 .agents/skills/minimax-h3
Codex 會從目前資料夾往上到儲存庫根目錄,在 .agents/skills 中尋找技能;用 /skills 或 $minimax-h3 呼叫。所有專案:~/.agents/skills。
git clone --depth 1 --branch main https://github.com/calesthio/OpenMontage.git /tmp/OpenMontage mkdir -p .cursor/skills cp -r /tmp/OpenMontage/.agents/skills/minimax-h3 .cursor/skills/minimax-h3
Cursor 啟動時會載入技能;在 Agent 聊天輸入「/」呼叫。Cursor 也會讀取 .claude/skills 與 .agents/skills。所有專案:~/.cursor/skills。
git clone --depth 1 --branch main https://github.com/calesthio/OpenMontage.git /tmp/OpenMontage mkdir -p .gemini/skills cp -r /tmp/OpenMontage/.agents/skills/minimax-h3 .gemini/skills/minimax-h3
任務符合時,Gemini CLI 會提議啟用技能,由你確認。所有專案:~/.gemini/skills。
安裝前請先檢視 SKILL.md 與相關檔案。點擊檔案即可預覽。
SKILL.md 內容
作者原文 · 唯讀MiniMax H3
MiniMax H3 is the Hailuo 3.0 family. The first-party API identifier is
MiniMax-H3; fal.ai exposes the family as hailuo-03, and Runway uses
hailuo3. Do not substitute one provider's identifier into another API.
Choose a route
| Route | Tool call | Execution |
|-------|-----------|-----------|
| MiniMax direct | minimax_video, model: "MiniMax-H3" | Hosted first-party v2 API; global or mainland-China region |
| fal.ai | minimax_fal_video | Hosted gateway; T2V, I2V, reference-to-video |
| Runway | runway_video, model: "hailuo3" | Hosted; 768P or 2K, 5–15 seconds |
| ComfyUI Partner Node | comfyui_video, model_family: "minimax_h3_api" | Hosted and billed in Comfy credits |
| ComfyUI open weights | comfyui_video, model_family: "minimax_h3_local" | Local GPU with official workflow and model stack |
For local ComfyUI, export the official workflow in API format and pass
workflow_json or workflow_path plus output_node. OpenMontage reports the
required diffusion model, Qwen3-VL text encoder, video VAE, and audio VAE; it
does not silently download large weights.
Operations and prompting
- Text-to-video: concrete ratio; do not use
adaptivewithout visual input. - Image-to-video: provide a first frame.
- First/last-frame: provide both images and describe the motion between them.
- Reference-to-video: images, videos, and audio can be combined. Audio needs at
least one visual reference.
Write prompts as subject + action + camera path + environment + lighting +
audio intent. MiniMax responds well to explicit camera direction. Keep the
requested motion achievable within 4–15 seconds and inspect native audio as
carefully as the image track.
Provider differences
The direct MiniMax v2 route currently outputs 2K and supports 4–15 seconds.
Runway's Hailuo 3.0 route supports 768P/2K and documents 5–15 seconds. Partner
Nodes require network access and credits. Only the open-weight ComfyUI route is
local/offline after all models are installed.
常見問題
minimax-h3 技能有什麼用途?
使用各種 API 和本地模型生成 4-15 秒的 2K MiniMax H3 (Hailuo 3.0) 影片。
如何安裝 minimax-h3?
打開本頁的「安裝」區塊並選擇你的代理 — Claude Code、Codex、Cursor 或 Gemini CLI:那裡有把技能資料夾複製到正確位置的現成指令。也可以複製「透過助理」分頁中的提示,讓助理自行檢查並安裝技能。
minimax-h3 適用哪些 AI 模型?
技能是純 Markdown 指示(Agent Skills 格式),因此 minimax-h3 適用任何讀取 SKILL.md 的代理:Claude Code、Codex、Cursor、Gemini CLI、GitHub Copilot、OpenCode 等。只有在任務符合技能描述時,模型才會載入指示。
minimax-h3 是免費的嗎?
此技能由 calesthio 公開發布,採用 AGPL-3.0 授權。技能本身免費 — 你只需向 AI 供應商支付實際使用的 token 費用。
安裝技能安全嗎?
技能是代理會遵循的指示,安裝前請閱讀 SKILL.md 及相關檔案。技能資料夾中的腳本只有在代理執行時才會運行 — 留意它的動作,不要授予多餘權限。
技能和 MCP 伺服器或外掛有何不同?
MCP 伺服器或外掛為模型提供新工具 — 會執行並回傳結果的程式碼。技能提供的是知識與流程:如何處理任務、要遵循哪些步驟、好的成果長什麼樣子。它不需要伺服器或 API 金鑰。