For about-kazakhstan.com I needed cover images for 83 articles. Some topics work with AI generation, others absolutely dont. Here is what I learned.
When to Use AI (FLUX/Runware)
- Landscapes, abstract scenes, atmospheric shots
- Historical illustrations (Kazakh Khanate, Silk Road)
- Nature scenes (mountains, steppes, canyons)
I use Runware API with FLUX.1 Dev (Nano Banana 2 model) at 1344x768px.
# generate-covers.py
response = runware.generate(
model="google:4@3", # Nano Banana 2
prompt="Vast Kazakh steppe at golden hour, mountains in distance",
width=1344, height=768
)
When to Use Real Photos (Pexels API)
- Cities, buildings, streets
- Food and cuisine
- People and clothing
- Anything where accuracy matters
AI generates plausible but wrong food. Kazakh beshbarmak from AI looks nothing like real beshbarmak. For Kazakh food articles, always use real photos.
// Pexels API for real photos
const photos = await pexels.photos.search({
query: "kazakh traditional food",
per_page: 5
});
The Rule
AI for vibes. Real photos for facts. Never use AI for cities, food, people, or cultural items. Your readers will notice.
Cost
- Runware: ~$0.01 per image
- Pexels: free (with attribution)
- Total for 83 covers: under $5
How do you handle blog imagery? Pure AI or mixed approach?
Top comments (0)