AI Prompts for Adjusting Content for Nano Banana

Run a baseline test prompt using only the core subject '[PLACEHOLDER]'. Analyze the AI's response to identify if the issue originates from the core subject or if it is influenced by additional descriptors. Provide a detailed explanation of your findings and any recommendations for further testing.

Provide a known-working reference prompt that successfully generates content similar to your desired output. Use the format '[REFERENCE PROMPT]' and explain how this prompt aligns with your goals. Additionally, analyze any differences between your prompt and the reference to identify potential issues.

Modify one parameter in your prompt to test its impact on the AI's response. For example, change the style preset to '[NEW STYLE PRESET]' and analyze whether this adjustment influences the AI's ability to generate the desired content. Provide a summary of your findings based on the changes made.

Generate four variations of the following prompt using '[PLACEHOLDER]' as the subject. For each variation, modify the wording or structure while maintaining the core idea, and analyze the differences in the AI's responses to understand which version is most effective.

Generate an image based on the description provided and compare it against the reference image. Identify specific elements in the generated image that differ from the reference and explain how these discrepancies may contribute to the refusal of the content.

Rephrase the following problematic descriptors in your prompt to make them more neutral or generic. For instance, replace 'dark and ominous' with 'mysterious and intriguing'. Provide at least three examples of descriptors that need rephrasing and suggest appropriate alternatives for each.

Incorporate negative prompt elements to clarify what you do not want in the image. For example, specify that the image should be created 'without violence or explicit content' and provide additional details about the themes or styles to avoid, such as 'no dark colors or gloomy atmospheres.

Switch to a different style mode that is less likely to trigger content refusal. Identify the specific style preset you want to use, and then provide the prompt you wish to enter after selecting '[ALTERNATIVE STYLE PRESET]' from the left panel.

Restructure the following prompt by placing the subject first and the style last. For example, transform it to the format '[PLACEHOLDER] in the style of [STYLE]'. Provide a revised version of the prompt that follows this new structure while maintaining its original intent.

Switch to a different model version in Nano Banana to improve the handling of your prompt. Check the settings and select '[NEW MODEL VERSION]'. After switching, test your prompt again and note any differences in the response quality.

Generate four new variations of the prompt '[NEW PROMPT]' to confirm that the adjustments made have successfully resolved the refusal issue. Each variation should maintain the core intent of the original prompt while incorporating different phrasing or structure to ensure clarity and effectiveness.

Run the adjusted prompt for the use-case of nano-banana-refusing-allowed-content and compare the output to the Stage 1 baseline. Identify specific improvements in both acceptance rates and overall quality, providing detailed examples of changes observed in the output.

Generate an image using the prompt '[HARDER SUBJECT]'. Ensure that the subject is more complex or challenging than previous attempts, and describe any specific elements or details that should be included to enhance the complexity of the image.

Lock the seed value to ensure consistent results across generations. Specify the seed value you want to use as '[YOUR SEED]' and explain how this will affect the output consistency in your results.

Check the output of your adjusted prompt at a different aspect ratio. Select the desired aspect ratio from the dropdown menu and analyze how the change impacts the generated content, noting any differences in quality or detail.

Create a reusable base template for generating acceptable content. The template should include the phrase '[PLACEHOLDER]' and be written in a neutral style, ensuring it avoids any explicit elements. Provide examples of how this template can be applied in different contexts.

Create a negative prompt bank that includes phrases such as 'without violence', 'no explicit content', and 'not disturbing'. Provide at least ten examples of how these phrases can be effectively integrated into various prompts to ensure content adheres to these guidelines.

Create a comprehensive checklist for ensuring style consistency in content creation. Include steps to verify the selected style preset, confirm that all descriptors align with platform guidelines, and provide examples of common inconsistencies to watch for.

Establish a detailed workflow for locking seeds in your prompts to ensure consistency across generations. Include the instruction to always add 'seed: [YOUR SEED]' in your notes, and outline the steps to verify that the seed is correctly applied in each prompt iteration.

Create a prompt using the following template: '[PLACEHOLDER], in the style of [STYLE REFERENCE], without [NEGATIVE ELEMENTS].' Ensure that the prompt is specific to the topic you want to address and clearly defines the desired style and any elements to avoid. Provide an example of how this template can be applied effectively.