Currently, the Qwen-Image LoRA is undergoing continuous training; the current LoRA is a test version.

Style Introduction:

"Atmospheric Narrative": Atmospheric light and shadow, featuring compositions that blend characters and environments seamlessly.

Qwen-Image Prompting Suggestions:

Clarity: Prompts should clearly and accurately describe the desired image content, avoiding vague or ambiguous phrasing. For example, if you want to generate a cat, do not just say "animal", but specify "an orange Persian cat lying on a soft white blanket, with blue eyes and a lazy expression".

Richness of Detail: Provide as many details about the image elements as possible, including the subject's features, actions, environment, lighting, colors, style, etc. For example, when generating a character image, you can describe the hairstyle, outfit style and color, facial expression, posture, as well as whether the background is indoor or outdoor, and if there are specific decorations or weather conditions.

Style Specification: If you have specific artistic style requirements, such as realistic style, cartoon style, traditional Chinese style, cyberpunk style, etc., explicitly specify them in the prompt. For example: "A traditional Chinese illustration of a woman wearing a light blue Ruqun dress, standing under a blooming peach tree, holding a round fan; delicate and beautiful art style with elegant, soft colors."

Logical Consistency: The content described in the prompt should be logical, and elements should complement each other reasonably. For example, avoid descriptions that contradict common sense, such as "a fish in space surrounded by flowing seawater".

Conciseness of Language: While ensuring complete information, keep the language concise and clear, avoiding verbose or wordy descriptions to prevent affecting the model's ability to extract key information.

Reprinted from: Xingmeng｜Qwen-Image｜Photography Style｜Atmospheric Narrative