Bilingual Text-to-Image Prompt Generator

Problem

Text-to-image models are sensitive to prompt structure and detail density. A short request such as “cute product poster” or “Chinese-style illustration” often produces unstable results, especially when lighting, composition, material, and technical specs are missing. This skill turns a rough visual brief into a reusable, inspectable prompt artifact for GPT-4o, Midjourney, DALL-E, and domestic platforms.

How It Works

• Requirement parsing: extract subject, style, use case, scene, and element constraints, such as e-commerce KV, brand logo, portrait photography, 3D rendering, or infographics.
• Template matching: choose a template by category, including photography, e-commerce, brand design, ink painting, movie posters, and cultural themes.
• Bilingual generation: build the English prompt around subject + style + lighting + composition + technical specs, then rewrite the Chinese version with the same level of detail.
• Variable handling: replace placeholders such as [BRAND NAME], [SUBJECT], [PRODUCT], and [color1], while avoiding leftover [UNDEFINED] tokens.
• Quality checks: verify completeness, negative-prompt fit, concrete style wording, and technical tags like 8K or hyper-realistic.

Boundaries

This is useful when designers, product teams, or engineers need to convert ambiguous visual requirements into cross-platform prompts. It does not replace reference images, brand guidelines, or human aesthetic judgment. For characters, logos, or complex narratives, still define exact subject features, ownership limits, and acceptance criteria.