The following steps are recommended to improve the generation of the Chivi AI:
- Use of specific spellsAvoid vague terms such as "good-looking" and describe details (e.g., "woman in red Chinese dress in a green landscape"), including keywords for scene, color, and style.
- Combined function use: Enable the "Negative Mantra" function on the generation page to exclude undesirable elements (e.g. "face distortion | background clutter").
- staged test: Generate a low-resolution sample with 1 credit, confirm the composition and then spend 5 credits to output a high-resolution version.
- The Case for Learning Communities: View the full spells and parameter settings of the best works in the creation community, and directly reuse validated and effective combinations of descriptions
Platform algorithms respond more accurately to the figurative description, it is recommended to build their own keyword library (such as "chapping | white | gold leaf effect" and other terms).
This answer comes from the articleQiwu: AI tool for generating Chinese aesthetic art paintingsThe
May not be reproduced without permission:AI productivity tools " How can I optimize my creation experience on the Qiwi AI to get a national style painting that better meets expectations?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil