OpenAI's image generator on a 400K context
GPT-5 Image generates images from text and reference material, and accepts file input alongside images. Its 400K context is far larger than the Gemini image models', which matters when the brief itself is long.
Try GPT-5 Image