qwen-image-2512
Adapt model-neutral still-image briefs into Qwen Image 2512 prompts for portraits, general images, exact visible text, character look development, locations, props, keyframes, and production stills.
SKILL.mdskills/qwen-image-2512/SKILL.md
GitHubQwen Image 2512 Prompting
Use the Qwen 2512 production prompt grammar while preserving project canon.
Input
Read one image brief plus canon and any named reference descriptions.
Workflow
- Classify the task as portrait, text-containing image, or general image.
- Write one continuous English description rather than a tag dump.
- Preserve proper names exactly.
- Add visual detail only when it is consistent with canon and useful to composition, lighting, material, atmosphere, or style.
- If visible text is required, quote every exact string and state placement, orientation, typography characteristics, color, and carrier. Do not translate or rewrite supplied text.
- If no visible text is required, explicitly state that the image contains no recognizable text when that constraint matters.
- For a character, use only identity, age, physical, wardrobe, and grooming facts established by canon. Do not invent ethnicity, age, or body traits merely to make the prompt longer.
- Aim for a concise but complete production description. Remove redundant quality words.
- Write the prompt under
04_generation/prompts/qwen-image-2512/<source-id>.md.
Output
SOURCE_ID: ...
MODEL: qwen-image-2512
PROMPT:
<single continuous image description>
CONSTRAINTS:
- ...
Done
The prompt preserves every hard visual fact and every exact text string from the source brief without adding conflicting canon.