Image description from the gpt-4-vision

[Simakvokka]

Along with the implemented images generation from description adding image text description would be great. With ability to add custom prompts.