General tips on how to use Scenario by Rodrigo
This is a basic guide for new users of general tips on how to use scenario.gg. These learned tips that I would like to share here.
Tips on Writing Prompts
Put the most important keywords first in your prompt, as the order of the words can make a difference in the result.
Learn about using Prompt Expressions on Scenario by clicking the following link to easily adjust the weights of your prompts:
Prompt Expressions
Start the prompt with a descriptive phrase about the image, and then include keywords separated by commas.
Set your guidance between 7 and 10, as it is very rare to need to increase beyond 10. In some cases, you may need to increase guidance up to 20, but keep in mind that this can cause overfitting.
For "sampling steps" start with a value from 50 to 100, when you get a result close to what you want, try going above 100 to experiment with different results.
Use keywords like "high quality," "high detailed," "beautiful eyes," etc., when you want to improve something specific. Synonyms can also be helpful.
Use negative prompts to provide specific corrections, and use prompt expressions to add even more emphasis if needed.
Tips for Using img2img:
Adjust the aspect ratio by clicking the icon at the bottom right
Removing the background can make a big difference, especially when generating a character.
Balance the influence of the image and the guidance settings to achieve the desired result. Instead of increasing both values together, try decreasing one and seeing if the other improves.
Colors can significantly impact the img2img result. One workaround is to change or paint the colors in a basic way in a photo editing program like Photoshop.
Tips for Creating Generators
Choose only 5 to 10 images with different characteristics. When creating a character generator, for example, choose images with varying proportions, poses, colors, and clothing. If desired, you can include some images with similar poses to maintain the same pose.
Test the generator by generating images without any prompts to see how it looks. Then, add keywords for fine-tuning.
I learned these tips through trial and error, as well as with the help of @emm, @mint, and @bubalu in the Discord community.
I hope this guide is helpful. If you have any other tips or suggestions, please let me know!
Follow me on twitter for more guides: https://twitter.com/rodrigon
Tips on Writing Prompts
Put the most important keywords first in your prompt, as the order of the words can make a difference in the result.
Learn about using Prompt Expressions on Scenario by clicking the following link to easily adjust the weights of your prompts:
Prompt Expressions
Start the prompt with a descriptive phrase about the image, and then include keywords separated by commas.
Set your guidance between 7 and 10, as it is very rare to need to increase beyond 10. In some cases, you may need to increase guidance up to 20, but keep in mind that this can cause overfitting.
For "sampling steps" start with a value from 50 to 100, when you get a result close to what you want, try going above 100 to experiment with different results.
Use keywords like "high quality," "high detailed," "beautiful eyes," etc., when you want to improve something specific. Synonyms can also be helpful.
Use negative prompts to provide specific corrections, and use prompt expressions to add even more emphasis if needed.
Tips for Using img2img:
Adjust the aspect ratio by clicking the icon at the bottom right
Removing the background can make a big difference, especially when generating a character.
Balance the influence of the image and the guidance settings to achieve the desired result. Instead of increasing both values together, try decreasing one and seeing if the other improves.
Colors can significantly impact the img2img result. One workaround is to change or paint the colors in a basic way in a photo editing program like Photoshop.
Tips for Creating Generators
Choose only 5 to 10 images with different characteristics. When creating a character generator, for example, choose images with varying proportions, poses, colors, and clothing. If desired, you can include some images with similar poses to maintain the same pose.
Test the generator by generating images without any prompts to see how it looks. Then, add keywords for fine-tuning.
I learned these tips through trial and error, as well as with the help of @emm, @mint, and @bubalu in the Discord community.
I hope this guide is helpful. If you have any other tips or suggestions, please let me know!
Follow me on twitter for more guides: https://twitter.com/rodrigon
Updated on: 04/05/2023
Thank you!