CompVis/stable-diffusion vs XingangPan/DragGAN

Updated 20:30 UTC.

CompVis/stable-diffusion has more stars (73,576) than XingangPan/DragGAN (35,692). In the last 7 days CompVis/stable-diffusion gained +4 stars and XingangPan/DragGAN gained -8. CompVis/stable-diffusion license: Custom license. Use at work: Check the license. Maintenance: Not maintained. XingangPan/DragGAN license: Custom license. Use at work: Check the license. Data as of 16 Nov 2026.

Stars gained since 4 Nov

  • stable-diffusion ▲ +2
  • DragGAN ▼ −18

By the numbers

Not legal advice
MeasureCompVisstable-diffusionLatent text-to-image diffusion modelXingangPanDragGANDrag points on a GAN image to edit it
Stars73,57635,692
Last 7 days▲ +4▼ −8
LicenseCustom license Check the licenseCustom license Check the license
Setup difficultyhardhard
MaintenanceNot maintainedUnknown

Which one fits

QuestionCompVisstable-diffusionXingangPanDragGAN
Fit for
  • People with a GPU of 10GB VRAM or more who want to sample the model
  • Researchers and developers studying text-to-image diffusion models
  • Researchers and developers exploring interactive GAN image editing
  • Users with a CUDA GPU who want to run the official DragGAN code
Not for
  • Users without a suitable GPU
  • Product teams needing a ready, safety-tested service
  • People wanting a one-click hosted tool with no setup
  • Editing real photos directly without a GAN inversion step