Hunyuan Designer

Problem

Design leads spend time on schedules, version chaos, and inconsistent junior output. The issue is not just prompt writing; it is losing seeds, finding final files, missing brand wording, and having no rule for reference images. This skill makes Tencent Hunyuan Image 3.0 calls more engineering-like: identify banner, poster, square, or scene image, select templates, prepend brand style words, set resolution, and record seed, prompt, and version.

How It Works

The workflow covers three scenarios:

• Fast generation: use scenario templates for first drafts, with common sizes such as 1024:512, 768:1024, and 1024:1024.
• Version control: maintain changes using v1.0 initial, v1.1 micro-tuning, and v2.0 major revision; reuse seeds and edit prompt fragments, or use previous image as --images reference for composition changes.
• Brand consistency: keep brand style words at the front, anchor outputs with a baseline seed and checklist, and reduce rework.

The script layer defaults to scripts/main.py for submission and polling, with --seed, --resolution, --images, and --logo-add as key parameters. Returned URLs should be saved within one hour.

Boundaries

It suits visual work with clear goals, batch generation, and version tracking, not brand strategy, rights review, or complex layout. When --revise is enabled, a fixed seed may not take effect; output resolution adapts to supported sizes and may differ from input; watermark is off by default and should only be enabled on request.