Features
Text-to-image generation from written prompts
Bilingual Chinese-English text rendering in images
Image editing via instruction-based Edit model
Multiple model intents: Base, Turbo, Edit, FP8
Adjustable inference steps, guidance scale, aspect ratio
Negative prompt support to avoid unwanted elements
Output in JPEG or PNG formats
Asynchronous generation with task polling
Credit-based usage with free daily credits
Public gallery option for generated images
Web-based, no local GPU required
Flexible rendering speeds: TURBO, BALANCED, QUALITY
API with generate and status endpoints
Prompt patterns for portraits, products, and scenes (Pro and above)
AI upscaling up to 6x lossless zoom
Nine specialized upscale models (Standard, High Fidelity, Low Res, Art & CG, Face Recovery, Wonder, Standard Max,
Wonder Model combines upscale, sharpen, denoise
Face Recovery for portrait enhancement
Art & CG model for digital artwork and illustrations
Text & Shapes model for scanned documents
Redefine model for heavily degraded images
Unlimited local rendering (files never leave computer)
Unlimited cloud rendering with 2-image concurrency
NeuroStream 2 for faster local AI performance
Plugins for Adobe Photoshop and Lightroom Classic
Standalone desktop app for macOS and Windows
Before/after preview with drag to reveal
Precision Update (March 2026) improved accuracy
