minimax_skill

Overview

This skill exposes the MiniMax API via curl for Chinese-optimized LLM chat, high-quality text-to-speech, and AI video generation. It provides concrete curl examples, model recommendations, and tips for streaming, async video tasks, and emotion-controlled TTS. Use it to integrate MiniMax features into shell scripts or CI workflows quickly.

How this skill works

The skill demonstrates authenticating with a Bearer token and calling MiniMax endpoints for chat completion, TTS, and video generation using curl. It covers synchronous and streaming chat, model choices (M1/M2 for reasoning, Text-01 for speed), TTS parameters (voice, emotion, pitch, bitrate), and async video workflows that return task IDs to poll. Examples include both text-to-video (T2V) and image-to-video (I2V) payloads and camera control syntax embedded in prompts.

When to use it

• Build Chinese-first chatbots or assistants with MiniMax-M1/M2 models
• Generate natural-sounding speech with emotion control for narration or IVR
• Create short videos from text prompts or convert images into animated clips
• Require streaming outputs for long-form reasoning or step-by-step explanations
• Integrate via shell scripts, CI pipelines, or simple serverless functions

Best practices

• Set MINIMAX_API_KEY as an environment variable and use bash -c '...' when piping to avoid lost env vars
• Choose M1/M2 with stream: true for complex reasoning and to receive intermediate reasoning_content
• Use Text-01 or speech-02-turbo for faster responses when quality tradeoffs are acceptable
• For video generation, handle the returned task_id asynchronously and poll for completion
• Embed camera commands in brackets (e.g., [Tracking shot, Pan left]) and limit to three simultaneous movements

Example use cases

• A Chinese customer support assistant that streams step-by-step problem diagnosis
• Batch TTS generation producing voiced assets with emotional tones for e-learning
• Automated short marketing videos from product descriptions using T2V models
• Image-to-video social posts that animate photos with subtle camera moves
• Serverless job that submits video tasks and updates status based on task_id polling