Customer feedback: No support for text-to-image generation even though some models (e.g., GPT-5.2) provide that feature