phy-fal-image-gen_skill

Overview

This skill generates professional images using fal.ai and BytePlus Seedream APIs. It supports single-image, parallel batch, and reference-guided generation for photos, posters, and other visual assets. Choose between cheaper, photo-real Seedream or feature-rich fal.ai models depending on cost, quality, and text-rendering needs.

How this skill works

Provide a text prompt and optional reference image URLs; the script auto-selects a backend or honors a user-specified model alias. It can run single requests, JSON-driven parallel batches, or structured JSON prompts for poster-style generation. Outputs are saved to dated output folders and printed as MEDIA: lines for gateway delivery or downstream ingestion.

When to use it

• Generate photoreal images from short or detailed text prompts with optional reference images.
• Produce posters or infographics that need high-quality text rendering using the NBP poster model.
• Run large-scale batches of visual assets in parallel for campaigns or A/B tests.
• Create images for social platforms with specific aspect ratios and resolution tiers.
• Quick prototypes where cost matters — default routes to the cheaper Seedream backend.

Best practices

• Specify model aliases when you need a particular backend or text-rendering behavior (e.g., --model nbp for poster text).
• Use references for consistent style or layout; Seedream supports up to 14 refs, NBP supports fewer.
• Pick a resolution tier based on use: standard for social, high for downloads, max for print.
• Use JSON structured prompts for poster layouts with NBP to leverage its JSON parsing strengths.
• Keep prompts concise but descriptive; include lighting and film references for predictable desaturation control per model.

Example use cases

• Create a travel poster with a JSON-structured prompt and NBP for crisp typographic rendering.
• Batch-generate 50 social images with mixed models and aspect ratios from a single JSON file.
• Generate a product photo series by supplying multiple reference images to Seedream for consistent color and lighting.
• Produce vertical story assets (9:16) for Xiaohongshu or Instagram Stories using the standard resolution for fast turnaround.