Dec 21, 2025
v1.2.5
Layered output support, new WAN v2.6 models, Flux 2 Max, and SVG paste
Models & workflows
New (12)
Qwen Image Layered (fal.ai) - decompose images into 1-10 RGBA layers
Qwen Image Layered (Replicate) - same layered decomposition via Replicate
Qwen Image Layered (RunningHub) - layered decomposition workflow
Trellis 2 - image-to-3D GLB model generation
WAN v2.6 Text-to-Video - multi-shot video with scene segmentation
WAN v2.6 Image-to-Video - motion synthesis with multi-shot support
WAN v2.6 Reference-to-Video - character consistency with @Video1/@Video2/@Video3
WAN Vision Enhancer - video enhancement with creativity levels
WAN 2.6 Text-to-Video (Replicate) - multi-shot generation via Replicate
WAN 2.6 Image-to-Video (Replicate) - motion synthesis via Replicate
Flux 2 Max - premium text-to-image with exceptional realism
Flux 2 Max Edit - advanced editing with up to 9 reference images
Archived (1)
Legacy Replicate FLUX models (available via Archived filter)
What's new
Layered output support: AI models that produce multiple layers (Qwen Image Layered) now show violet layer badges, support AE export as stacked layers, and include fullscreen layer navigationWatch
Improvements
SVG clipboard paste: Paste SVG content directly from clipboard as .svg files with automatic security sanitization
Fixes
Fullscreen audio preview now shows waveform and playback controls