Hi territ61090028,
I want to clarify how the Generate Image capability works so we can get better results more efficiently.
The image generator does not function as a conversational LLM. Each prompt is processed independently, which means it doesn’t retain context or “learn” from previous generations. Because of this, iterating conversationally within it (e.g., refining step-by-step across multiple prompts) won’t produce consistent or improved results.
For best outcomes:
- Treat each generation as a self-contained request
- Include all necessary details, constraints, and context directly in the prompt
- Avoid relying on prior outputs to inform the next one
- Refer to this article to learn how to write an effective text prompt
If you're looking for a more conversational workflow, where you can iteratively refine, guide, and improve outputs over a single thread, you’ll get much better results using AI assistance (chat-based tools) instead. Those are designed to maintain context and support progressive refinement.
Using the right tool for the right type of interaction will save time and significantly improve output quality.
Let us know if you have any other questions.
Hajar