WHAT IT DOES
Features
in detail.
- Photorealistic and stylized images
- Prompts in Russian and other languages
- Choosing compatible LoRAs
- Hires Fix and separate upscaling
Build the scene from the general to the specific
Start with the space and main event: a quiet station before dawn, a sculptor's workshop or a small rooftop garden. Then define viewpoint and lighting. Add details once the overall image is clear. In an atmospheric illustration, weather and color may matter more than object count; for a character, silhouette, clothing and expression take priority.
Do not combine incompatible directions without a reason. Simultaneously requesting documentary photography, flat collage and oil painting makes the result hard to judge. Choose a foundation, then deliberately add a second technique. Keep the subject constant when comparing styles so you can see how style changes the image.
Account for YesAI's implementation
SDXL is a separate Stable Diffusion branch. In YesAI, this mode uses text prompts; references and masked editing are unsupported. To process an existing picture, choose a catalog editor.
Choose LoRAs designed for SDXL. Explore available settings first, then run a clear experiment: one subject, one change. This separates the prompt's influence from the additional model's influence.
Evaluate the image against the task
For a game background, check character space and readable depth. For a slide, check room for the headline. For a resource card, check that its main object remains recognizable at a small size. A vivid image can fail in a specific layout; define empty zones early and avoid cluttering the center with secondary details.
Hires Fix and separate upscaling are available enlargement methods. Use them after checking shape and content. If a specific area must change, a textual refinement in a new generation may produce another image rather than a precise edit of the previous one. Choose a suitable editor when preserving the source matters. The SDXL form below sends only a new-image description to YesAI.
Suitable tasks
- Background for a visual story
- Landscape with a particular mood
- Character concept
- Illustration for a website or presentation
Things to consider
- SDXL in the current YesAI mode has no input references or inpainting.
- Compatibility with another branch's LoRA is not implied.
- Editing an existing photograph requires another tool.
FROM AN EXAMPLE TO YOUR OWN IDEA
Where to start a prompt
Choose an example and insert your own details.
Story background
An empty station in a small seaside town before dawn. Platform-level view, tracks receding into the distance, soft mist and a few cold lights. A painterly background for a visual novel. Leave the center free for a character, with no text or billboards.
Slide illustration
An isometric rooftop garden on a low-rise building: paths, greenery and a small seating area. Simple, readable volumes, a white background, blue and coral accents. Leave room for a headline on the left. No people or lettering.
Character
A traveling female botanist with a field bag and notebook, comfortable clothing and a clear, recognizable silhouette. Full length, neutral light background, digital illustration with soft shadows. No weapons, logos or tiny text.
Workflow
- 01
Write a complete description of the intended frame.
- 02
Open SDXL and check the selected add-ons' compatibility.
- 03
Evaluate the composition and upscale if needed.
A FEW MORE ANSWERS
Questions about SDXL
Does SDXL support photographs as references?
Input references are unsupported in the current YesAI mode. Create a new image from a description or choose another tool to process photographs.
Can I use a Stable Diffusion LoRA?
Check the specific branch. A LoRA for Stable Diffusion 1.5 is not automatically compatible with SDXL; choose versions designed for SDXL.
How can I create a series of similar illustrations?
Keep descriptions of light, palette and technique, changing only the subject. This helps retain direction, but each result still needs comparison with the rest of the series.
