Advanced Guide: Modify Game Generation Results via Prompt

1. Stably modify the generated results in multi-round conversations

After initially generating content, how can we talk to Boo to make our game look next level?

1. First affirm, then advance

Prevent ​Boo​​ from mistakenly thinking that a full ​rewrite​​ is needed

First, affirming the existing results can clearly convey:​​ The current achievements are retainable and only require local ​optimization​**.**

This approach has two advantages:

• Boo will continue to follow the current context and design direction
• Modifications are more focused and will not overturn the existing structure

Example:

• "You've done a great job. Now we just need to adjust the display issue of the central tower."

2. Describe the visible issue

Avoid ​Boo​​ self - interpreting in the wrong direction ​

Instead of letting Boo first judge "why it's wrong", directly point out: what problems are currently occurring on the screen, and what state you want it to be in.

This can reduce Boo's self-guessing of implementation logic, thereby aligning with the modification target more quickly.

Example:

• "There are two backgrounds during the game. Check and remove the bottom background, then adapt the top background to the screen."
• "Currently, the central tower is not fully displayed, with a blue square blocking it. Remove this blue square."

3. It is best to change only one dimension at a time

Prevent modification ​divergence​​ and problem stacking

When Boo is modified, it will by default perform "reasonable linkage" on related content.

Clearly defining what cannot be changed and what is the only modification point in this round can significantly improve stability. This is especially applicable to modifications that are prone to mutual influence, such as layout, proportion, and perspective.

Example:

• "Do not change the position of the ice cream ball, only change the position of the decorations."
• "This step only adjusts the perspective distance without changing the layout and dimensions."
• "Remove A, retain B, keep C unchanged, and only adjust D"

4. Reuse the parameters in the previous step

Maintain logical continuity and avoid repeated trial and error

When Boo has already provided the previous set of parameter suggestions, directly making fine-tuning on its basis is more likely to yield stable results than providing a completely new set of values.

Example:

• "You listed the minimum radius as 50px in the previous step, change the minimum radius to 20px."

2. When should we start over instead of continuing?

During the use of Combos, many users will encounter such a situation:

has gone back and forth with Boo for many rounds of revisions, but the result has always been "not quite right." Fixing one part seems to lead to new issues.

It's not necessarily that your description is not detailed enough, nor is it that Boo is "disobedient". ​ Instead, the current result is no longer suitable for correction through local modification. ​

The core criterion for determining whether to "start over"

If the modification is about "details", it can continue; if it is about "direction", it is recommended to start over. ​

1. Situations suitable for "Continue Editing"

Presentation and Layout Layer

• Element size is inappropriate
• Spacing is too large or too small
• Occlusion, misalignment, and incomplete display

Example:

• "The central tower is blocked by a blue square, please remove this block." - "The spacing between the ice cream creatives is too large, please move them closer together."

Local art or creative issues

• The style is generally correct, but some elements are inconsistent
• Selected state / Inconsistent style after placement

Example:

• "The images of the three types of towers are different when selected, but their styles are the same after placement. Please distinguish them separately."

Rule details and fine-tuning of player behavior

• The operation process is not smooth
• Feedback is not obvious
• The rhythm needs adjustment

2. Suitable for "starting over" situations

The game objective or experience direction is unclear

You find yourself constantly adding "Actually what I want is..." and "It should be more like..."

For example:

• The game types described before and after are inconsistent
• The gameplay changes from casual to strategic, or vice versa

This indicates that: ​ the initial experience target layer was not clear enough ​

Core gameplay or rules have changed

For example:

• Originally a synthesis gameplay, now we want to change it to tower defense
• Originally it was a single-screen operation, and now we want to add multi-region logic

This type of modification will affect the entire system structure and is not suitable for superimposed adjustment to existing results.

The overall art style needs to be completely overhauled

If you find yourself saying:

• "This content style is actually not quite right"
• "Let's try changing to another style"

That usually means: the art description layer itself needs to be rewritten

3. Starting over ≠ Starting from scratch

"Starting over" is not a failure, but a higher-quality beginning.

When re-describing, you can:

• Retain the settings that have been verified as effective
• Identify which points are "must-be-correct this time"
• Avoid overloading the first round with too many corrective descriptions

Many high-quality results precisely stem from a second, yet clearer, starting point.