Initially, I used OpenAI API for image generation, but I noticed that costs could spike sharply due to the peculiarities of the API's operation, leading to dozens of dollars in daily expenses.
Now I use replicate.com — the image quality is excellent, and the cost per image is less than 1 cent if there are no requirements for large size. Although images with people can be a bit glitchy, this API has other models to work with them.
I currently use the black-forest-labs flux-schnell model to create abstract covers and just launched a mass regeneration script on my openclawradar.com. I will discuss prompt optimization in a separate post.
