flux2-flash_skill

Overview

This skill generates and edits images using the Flux 2 Flash model (fal-ai/flux-2/flash). It automatically switches between text-to-image and image-editing modes based on whether input image URLs are provided. The model emphasizes photographic realism, improved text rendering, and native editing capabilities. It supports multiple image sizes, guidance control, and reproducible seeds.

How this skill works

You submit a task with model_id fal-ai/flux-2/flash and parameters via the submit_task API. If image_urls are omitted the skill runs text-to-image generation; if image_urls are present it runs image editing on up to four input images. The service returns a task_id which you poll with get_task to check status (pending, processing, completed, failed) and retrieve results.

When to use it

• Generate photorealistic images from text prompts
• Edit existing images (remove/add/replace elements) by providing image URLs
• Create images with legible on-image text thanks to improved text rendering
• Produce multiple variations by adjusting num_images or seed
• Quickly prototype scene, lighting, and composition changes

Best practices

• Provide detailed prompts describing scene, lighting, composition, and desired style for more accurate results
• Use explicit action verbs in edit prompts (Remove, Add, Change, Replace) to get predictable edits
• Set image_size to match your target aspect ratio and resolution (square_hd for 1024x1024)
• Control fidelity to the prompt with guidance_scale (0–20) and fix outputs with seed for reproducibility
• Enable prompt expansion when you want the model to enrich brief prompts

Example use cases

• Create a 16:9 landscape sunset with sailboats using a detailed scene prompt
• Remove an unwanted object (e.g., hamburger patty) from a provided product photo
• Generate packaging mockups with legible brand text rendered on the image
• Combine or compare up to four images in an edit workflow to produce a composite
• Produce multiple variants of a character or scene by changing seed and num_images