Fable
2Flow matching diffusion transformer model
About Fable 2
Fable 2 is a unified multimodal Flam generation model built on a flow matching diffusion transformers with support for image, audio and video as reference
Transform an idea into immersive visuals with exceptional motion stability, audio generation & full control over transparency, lighting, shadow and camera movement.
3rd%2520fold.webp&w=750&q=75)
Video models produce opaque, flat footage. Our dual-decoder architecture treats color and transparency as inseparable creative decisions. Its trained with edge-aware supervision to produce the crisp, compositable boundaries, and fine-tuned on curated datasets emphasizing dynamic motion, fine structural detail, and strong foreground presence
The Science of Compositing
Dual-Decoder Architecture
Color and alpha channels are processed together. Ensures temporal consistency across frames without any flicker
Edge-Aware Precision
Output smooth boundaries between transparent and opaque objects, no jagged edges
Optimized alpha channel
Produce alpha value between 2 to 98 percent with natural light blending. Easily create transparent particle and fluid effects