minimax-image-genlisted

# MiniMax Image Generation Skill > **What this skill does — generates images.** > Given a text prompt (or a prompt + reference image), this skill calls the > MiniMax image model and returns an actual image file on disk. It is the > skill to load whenever the user asks for "画一张", "生成图片", "做个海报", > "出一张配图", "draw me a picture", "generate an illustration", "create a > cover / logo / character sheet", or any visual deliverable. The skill wraps the official MiniMax endpoint: ``` POST {API_BASE_URL}/v1/image_generation ``` It ships: - a reusable Python client (`scripts/image_client.py`), - parameter validators (`scripts/validators.py`), - a CLI entry point (`scripts/generate_image.py`) — the **primary** interface, invoked via `bash` when the skill is in scope. ## Capabilities (when to use this skill) Use it whenever the user wants to **generate an image**, including: - **Text-to-image (文生图)** — turn a written description into a picture. - **Image-to-image with a subject reference (图生图 / 角色一致性)** — keep a character / object looking the same across new scenes. - **Multiple variations** — `--n 2..9` for "几张 / 多给几个 / a few options". - **Deterministic regeneration** — fix a `seed` for reproducibility. - **Specific aspect ratios** — `1:1`, `16:9`, `9:16`, `4:3`, `3:2`, `2:3`, `3:4`, `21:9` (image-01 only). - **Either URL or base64** output; the CLI downloads URL outputs to disk before they expire (~24 h). ### Trigger phrases (load this skill when you see these) - 生图 / 文