Generate Images with Fal
Promptfoo config comparing fal.ai's fast FLUX Schnell and high-fidelity FLUX Dev image models across seven artistic styles.
Why it matters
Leverage the Fal platform to generate high-quality images from text prompts. This asset provides a streamlined way to integrate image generation capabilities into your applications.
Outcomes
What it gets done
Define image generation parameters.
Execute image generation requests via Fal API.
Retrieve generated images.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/pfoo-fal-image-generation | bash Overview
Fal Image Generation
This Promptfoo example compares fal.ai's fast FLUX Schnell and high-fidelity FLUX Dev image-generation models on the same seed across seven artistic style prompts, from Van Gogh to Japanese woodblock print. Use it as a template when comparing fal.ai's FLUX Schnell and FLUX Dev models for speed versus fidelity across different artistic styles.
What it does
This Promptfoo config compares two fal.ai FLUX image-generation models: FLUX Schnell (4 inference steps, guidance scale 3.5, tuned for speed) and FLUX Dev (28 inference steps, guidance scale 7.5, landscape_4_3 aspect ratio, tuned for fidelity), both pinned to the same seed (42) for a reproducible side-by-side comparison. Seven test cases vary a shared prompt template's style, subject, mood, and quality variables across a wide range of artistic styles - Van Gogh, Picasso cubism, Monet impressionism, Studio Ghibli, cyberpunk digital art, Renaissance, and Japanese woodblock print - each paired with a distinct subject: a cyberpunk city, a chess-playing robot, a dragon, a floating castle, a neon street, an astronaut, and a giant octopus.
When to use - and when NOT to
Use it as a template for comparing fal.ai's fast (Schnell) versus high-fidelity (Dev) FLUX image models on the same prompt structure and seed, across a range of artistic styles. Do not use it if you only need one FLUX variant, or if you're not using fal.ai.
Inputs and outputs
Input: the YAML config - a shared prompt template with style/subject/mood/quality variables, and two FLUX provider configs with their own inference-step and guidance-scale settings. Output: Promptfoo's evaluation report showing the two models' generated images side by side for each test case - the config defines no pass/fail assertions, so review here is visual rather than automated scoring.
Integrations
Uses Promptfoo's fal:image provider to call fal.ai's FLUX Schnell and FLUX Dev models directly.
Who it's for
Teams choosing between fal.ai's FLUX Schnell and FLUX Dev models who want to compare speed and fidelity tradeoffs across a range of artistic styles before picking a default.
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.