hy-image-generation_skill

Overview

This skill integrates Tencent Cloud Hunyuan Image Generation 3.0 to produce high-quality images from text or from text+reference images. It offers a one-step workflow that submits a job and polls for results, plus separate scripts to submit and to query by JobId. The tool supports prompt rewriting, seed control for reproducibility, and optional reference images to guide style and composition.

How this skill works

The primary script runs a submission then polls the Tencent Cloud API until the image is ready, returning a temporary result URL. You can instead submit a job only to receive a JobId, then poll later with the query script. The skill automatically handles dependency installation, reads credentials from environment variables, and enforces resolution and reference-image constraints.

When to use it

• You want a simple, one-step generation that returns a usable image URL quickly.
• You need to submit a job now and poll later or from another system (use submit_job.py + query_job.py).
• You already have a JobId and want to check status or retrieve results.
• You want guided generation using up to 3 reference images for style or composition.

Best practices

• Use the recommended main.py one-step script for most user-facing requests to ensure zero-interaction execution.
• Provide clear, specific prompts; leave prompt rewriting enabled unless you will supply an optimized prompt.
• Limit reference images to 3 and use supported formats (jpg/jpeg/png/webp) with base64 size ≤10MB each.
• Respect resolution constraints: width/height in [512,2048] and width×height ≤ 1,048,576 pixels.
• Keep API keys in environment variables; never hard-code credentials in scripts.

Example use cases

• Generate a square product render for a landing page with main.py and default 1024:1024 resolution.
• Create variant art by submitting a job with a fixed seed to reproduce results later.
• Guide an illustration by supplying 1–3 reference image URLs that the model should follow in style.
• Submit many prompts in batch: use submit_job.py to collect JobIds and poll asynchronously with query_job.py.
• Request high-fidelity landscape or portrait images with custom resolution such as 1024:768 or 768:1024.