Xiaohongshu Food Note Maker

The Specific Problem It Solves

Creating high-quality food notes on platforms like Xiaohongshu requires a series of coordinated images (e.g., hero image, main dish close-up, detail shots, and ambiance) along with titles, copy, and tags. Manually producing this content is time-consuming and difficult to maintain visual consistency, especially when source materials (photos or concepts) are incomplete or fragmented.

How the Skill Works

This skill automates the transformation of a dish or restaurant visit into a complete four-page visual story. Its core capabilities include:
- Sequence Generation: By default, it creates a vertical 3:4, 2K resolution four-page image set delivered in order: hero image (with title safe area), main dish close-up, texture/detail shot, and dining ambiance.
- Image Processing Routes: It supports three technical routes:
- beatra.images.transform: Used when a source food photo is available, performing image-based transformations.
- beatra.images.generate: Used for pure dish or visit concepts, generating images via text prompts.
- beatra.images.edit: Used for focused refinement of approved pages to enhance details.
- Copy Accompaniment: Based on user-provided facts (e.g., dish details, restaurant information), it generates title inspiration, copy rhythm, and tags without fabrication.

Key steps follow a standard workflow:
1. Plan Food Note Card: Define the food subject, scene, target audience, style vocabulary, and reference images.
2. Select Processing Route: Choose the appropriate route based on input type (source photo or concept).
3. Prepare Prompts: Design the four-page prompts as a cohesive visual family to ensure consistency.
4. Confirm and Execute: After user confirmation, assign a stable client_request_id and submit the task, polling via beatra.tasks.get until completion.
5. Deliver Results: Return images in order, along with copy proposals and actual billing facts (e.g., artifact links, task IDs).

Applicable Boundaries and Considerations

• Input Requirements: The minimum input is a source food photo or a specific dish/visit concept; only queries that would substantially change the outcome are asked.
• Payment Confirmation: Planning and copy writing are free, but explicit confirmation is required before any image generation or refinement, freezing all parameters (e.g., subject, source image, prompts). Changing these constitutes new work and requires new confirmation.
• Model Compatibility: Uses model: "auto" by default, but before fixing a model, read beatra.models.list and check if the real-time model card supports sequence delivery. If not, present alternative options.
• Execution Dependency: All remote operations must use the bundled scripts/mcp_client.py; REST/OpenAPI cannot be used as a fallback.
• Scope Boundaries: Applies to food notes; for generic carousel images use zhongcao-carousel-maker, and for standalone covers use zhongcao-cover-maker.