Cinematic Fashion Editorial Shoot Generator

Problem

Turning one portrait photo into multiple editorial concepts and final images often gets stuck on weak visual variation, face drift, and missing review artifacts. This skill targets portrait shoots, magazine cover pre-generation, and visual marketing tests. It converts a source photo into three distinct fashion concepts, then produces three face-consistent cinematic hard shots.

How it works

The workflow is asynchronous. It starts with input compliance filtering, then extracts facial structure, feature proportions, hairstyle, and temperament tags from the source photo. It generates three independent editorial plans, including scene, wardrobe, pose, composition, lighting, and a professional English prompt. The model then generates images from those plans with the source photo as reference, and each final image receives a compliance review with a risk score and result code.

Key steps:
- Input: img_url is required; themes is optional; aspect_ratio can be configured.
- Execution: real calls use execute_task.py to submit a task and receive only a task_id; then query_task.py polls until completion.
- Output: editorial details, image URLs, review codes, and an HTML page for previewing images and concept data.

Limits

Without an appKey, the skill runs in simulation mode and returns one preview payload without the real async scripts. Formal runs require a configured ShowAPI showapi_appkey. Real tasks are asynchronous only. If review returns needManualReview, manual review is recommended. The source photo must be publicly accessible, and the output is better suited to concept testing, cover pre-visualization, and asset shortlisting rather than final commercial delivery without review.