Step 1: Select and mix srefs

Select the first sref

• Randomly discover a sref via sref, e.g. -sref 944260837

• This sref is bright and colorful, perfect for scenes where you want your image to be colorful.

Select the second sref

• Choose a soft sref, eg -sref 3755864991.

• This sref resembles a pencil drawing in color and texture.

• Mix this soft sref with the previous stronger sref to get a "in between" effect

Mixing Effects

• Mixing these two srefs gives an effect somewhere between strong color and soft texture.

Step 2: Adjust text hint and sref weight

Adjust text prompts

• Text cues have a huge impact on the final image.

• For example, after adding “photo” to the hint, the sref of the pencil drawing stands out more.

Adjust the weight of sref

• You can increase or decrease the weight of sref to control its influence in the image.

• For example, the weight of colored srefs was increased by adding "::2" (the weight was doubled), but it was found to have a poor effect.

• I then tried "::1.5" or "::1.25" and got a better result.

Step 3: Optimize the image using style references

• Once you have images you are happy with, you can add them as style references to further refine your results.

• Here, I added an image as a style reference, these are good.

Summary

• Mixing different srefs can produce unique image effects.

• Adjusting the weight of sref can fine-tune the visual effect of the image.

• The choice of text prompts will also significantly affect the final results. It is recommended to start with neutral prompts and then adjust them step by step.

• Once you find a satisfactory effect, use it as a style reference to further optimize the image quality.

• These methods are not fixed, you can try them as much as you want and see what surprising results you get.

👍🏼

Customize your server to fit your exact requirements