✦
G-star Studio
text-to-image · DAGroup-PKU/G-star · native 2× output
Workflow canvas →
Model card ↗
Prompt
Size (code grid × 32)
Seed
Randomize
Sharper ×0.5 output (paper-style Lanczos)
Enhance prompt (0.8B rewriter)
Sampler
Guidance (CFG)
Top-k
Top-p
Temperature
Refiner steps (lower = much faster)
Generate
Try
Your image will appear here.
~1–2 minutes on ZeroGPU; you get a replay of it emerging when it lands.
Queued
▶ Replay generation
Download
Reuse prompt & seed