Upload your source image and keep the direction grounded
变体 lets you start from a clear source image, so the output stays anchored to your subject, composition, and visual intent instead of drifting into an unrelated result.
变体
先挑一个更合适的起点，明确最终画面方向，再用更清爽的 Mango 风格 变体 工作流直接进入生成。
变体
Examples
Ref
Result
Court Angle Sheet
Create stunning AI images with 变体
查看最近生成结果、整理收藏，并把当前放大类流程集中在更干净的 Mango 工作台里完成。 Upload a source image, refine the visual direction, and generate cleaner images in a workflow that stays close to the look and structure of your original input.
Add a prompt to guide style, mood, and details
Describe the lighting, texture, background, or styling changes you want while preserving the structure of the uploaded source image. This makes it easier to iterate without rebuilding from scratch.
Choose the model that fits the result you need
Switch between available models to optimize for realism, speed, or stylization, then compare multiple generations until the final image matches your target direction.
How to use 变体
Source
01
1. Upload a clear source image
First, upload a clean and readable source image. This gives 变体 a stable visual base so it can preserve the key subject, composition, and structure before applying edits.
Prompt
Create a premium 2x2 multi-angle variation sheet from the uploaded reference image while preserving identity, outfit, lighting, and scene context.
02
2. Describe the result you want
Next, write a concise prompt describing the style, lighting, mood, composition, and specific changes you want. The clearer the instruction, the easier it is to steer the final result in the right direction.
Result
03
3. Generate, review, and refine
Finally, generate the result, compare the outputs, and refine the prompt if needed. Repeat until the final image reaches the quality and direction you want.
FAQs about 变体
What kind of reference files work best for 变体?
Source images with a clear subject, readable lighting, and a stable composition usually produce the most consistent results.
Can I guide the output with my own prompt?
Yes. You can upload a source image and still describe the styling, mood, detail, or background changes you want in the final output.
How do I get more consistent images?
Use a cleaner source image, keep the prompt specific, and stay focused on one visual direction instead of combining too many conflicting instructions.
Can I try multiple styles from the same input?
Yes. The workspace is built for iteration, so you can reuse the same source image and quickly generate several alternate directions.
Start creating with 变体
Upload your source image, refine the prompt, and generate polished AI images in one workspace.