Creating Images For The Paid247 Website

https://share.descript.com/view/yBTvxR9pF8C (opens in a new tab)

Deep-Dive: AI Image Generation Loop

How I move from blank page to on-brand visuals in under an hour.

1. Pin the Mood First

• Unsplash cache — drag 20–30 reference shots into one folder.

2. Image Training

• Train https://fal.ai/models/fal-ai/flux-lora-portrait-trainer (opens in a new tab) on the selected images

3. Write the Prompts

I open ChatGPT (model o3) and feed it it info on the brand and the content and ask it create prompts.

Template I reuse

<lora:FluxPortraitTrainer:1>, {subject}, {camera action}, rich {lighting adjective} light, brand colours weaving through scene, shallow depth, 35 mm DSLR, f/2.2 --ar 9:16

3. Generate → Curate in One Pass

Create images in https://fal.ai/models/fal-ai/flux-lora (opens in a new tab)

4. Feed into Sora & Framer

• Sora — drag three portrait JPGs into a storyboard row; let it auto-animate.

• Framer — use object-fit: cover; on image blocks so mobile crop stays tight on eyes.

• Verify that each page section’s background colour matches the dominant colour in its paired image (red section + red-tinted shot, etc.).