AI Prompts for Ensuring Negative Prompts Work

Generate a new prompt in VEO that focuses solely on the core subject: '[PLACEHOLDER]'. Analyze the output to determine if any negative elements are associated with the subject description, and provide insights on how these elements may affect overall performance.

Run the following reference prompt that successfully excludes negative elements: [REFERENCE PROMPT]. Compare the output of this reference prompt with the output of your original prompt to identify the specific variable that is causing the issue. Provide a detailed analysis of the differences observed and suggest potential adjustments to your original prompt.

Modify one parameter in your current prompt to test its impact on the output. For example, switch to a different model version or adjust the style setting. After making the change, analyze the results to determine if the issue you are experiencing is related to the specific configuration you altered.

Generate four variations of the following prompt: [CURRENT PROMPT]. Ensure that each variation alters the phrasing or structure while maintaining the core intent. After generating the variations, analyze the outputs to identify if the negative elements are consistently present across all versions.

Generate a video based on the prompt provided and compare it to the reference image or description. Identify specific instances where unwanted elements appear in the video, and detail how these instances differ from the reference to pinpoint the areas of drift.

Incorporate explicit negative prompts into your description by including phrases like 'without [UNWANTED ELEMENT]'. Provide at least three examples of how these negative prompts can be effectively used to clarify what should be excluded from the output.

Rephrase the following prompt to ensure it clearly excludes negative elements: "a bright scene". Specify the unwanted element by replacing [UNWANTED ELEMENT] with a specific descriptor, such as "a bright scene without shadows" or "a bright scene without clutter". Provide at least two alternative phrasings that maintain the positive focus while excluding the negative.

Switch to a different model version in the settings and evaluate how it handles negative prompts compared to the current version. Identify any improvements in the model's ability to respect exclusions and provide specific examples of changes in output quality.

Restructure the following prompt by placing the subject at the beginning and the style at the end: [PROMPT TEXT]. Explain how this change will help the model focus more on the core subject rather than on stylistic elements. Provide an example of the restructured prompt.

Revise your negative prompts to incorporate stronger descriptors that clearly convey your intent. For example, use phrases like "avoid [UNWANTED ELEMENT]", "exclude [UNWANTED ELEMENT]", or "do not include [UNWANTED ELEMENT]". Provide three examples of how these changes improve clarity and effectiveness in your prompts.

Generate four new variations of the adjusted prompt that eliminate the unwanted elements. For each variation, compare the output to previous generations to ensure that the negative aspects are no longer present.

Run a comparison of the current generation against the Stage 1 baseline generation. Identify any negative elements that are still present and provide a detailed analysis of whether they have been successfully excluded, including specific examples and suggestions for further improvements if necessary.

Test your prompt adjustments by providing a more complex subject that typically brings out unwanted elements. Specify the subject you want to use, and explain how it challenges the current prompt's effectiveness.

Run the adjusted prompt with a locked seed to ensure consistency in the output. Explain how locking the seed affects the stability of the results across multiple generations and provide examples of any variations observed.

Generate a video using a different aspect ratio than the original. Verify whether the negative prompts are still being respected in this new setting, and analyze if the issue is related to the specific aspect ratio used.

Create a base template for excluding elements in VEO. The template should follow this structure: '[SUBJECT], [STYLE], without [UNWANTED ELEMENT]'. Provide examples of how to modify the [SUBJECT], [STYLE], and [UNWANTED ELEMENT] for different scenarios to ensure clarity and usability.

Create a comprehensive bank of negative prompts that can be utilized in future projects. Include examples such as 'without [UNWANTED ELEMENT]' and provide at least five additional negative prompts that address common issues or undesired elements in various contexts. Ensure each prompt is clear and actionable for effective use.

Establish a workflow for locking the seed to ensure consistency in your outputs. Always include the instruction 'lock seed to [YOUR SEED]' at the end of your prompt to maintain reproducibility in your results.

Create a comprehensive checklist for maintaining style consistency in prompts. Include specific steps such as confirming that negative prompts are included, checking the order of prompts, and ensuring uniformity in tone and language throughout. Additionally, suggest methods for reviewing and revising prompts to enhance clarity and coherence.

Create a guide for structuring prompts effectively using the format: '[SUBJECT] [STYLE] without [UNWANTED ELEMENT]'. Provide examples for each component to illustrate how to apply this structure in various contexts, ensuring clarity and precision in future prompt generations.