تخطٍّ إلى المحتوى
أطلس الذكاء الاصطناعي

الصورة إلى صورة

إعادة توليد صورة مشابهة انطلاقًا من صورة

توليد الصور وتحريرهامبتدئ #19
إدخالصورةصورة

يُعرض النص الكامل باللغة الإنجليزية؛ وقد تمت ترجمة العنوان والملخص.

ما الذي تعنيه هذه القدرة

Takes an image and outputs a new one that resembles it in content but is rewritten in style or detail. The knob controlling similarity is usually noise strength: less noise stays close to the original, more noise approaches fresh creation. Unlike image editing it does not require a text instruction naming what to change; it restyles or reinterprets the whole image.

كيف تُنفَّذ تقنيًا

The method encodes the input into latent space, adds noise for a chosen number of steps, and denoises back from that point, so the output keeps the structure while taking on a new style. Finer control comes from stacking conditioning branches: edges, depth, pose or a reference style each enter as extra conditions, which is especially common in tasks such as completion and line-art colourisation.

منتجات تمثيلية

6

المؤسسات ذات الصلة

الاستخدامات الشائعة

  • Style transfer and photo stylisation
  • Sketch and line-art colouring and completion
  • Iterating from rough layouts to renders
  • Series of variants on one subject

كيف يُقاس مدى جودتها

FID
Distance between results and the target distribution
LPIPS perceptual distance
Perceptual difference from the input image
Structural consistency
How well contours and subject placement follow the input

الحدود والصعوبات

  • Balancing structure retention against rewriting is hard; the same settings behave differently across images
  • At higher strength the content drifts, and faces or text are the first details lost
  • Wholesale redrawing wrecks layout, so posters and UI screenshots cannot be preserved as-is

المفاهيم الكامنة وراءها