Decart Unveils Lucy Edit Dev: Revolutionary Text-Guided Video Editing

    
        By vramkickedin    
     | 
    
            October 2, 2025 at 1:45 pm        
    
     | 
    
        2 min read

# Lucy Edit Dev: Revolutionizing Text-Guided Video Editing

Open-Source AI Breakthrough in Video Manipulation

Decart AI has launched Lucy Edit Dev, the first open-weight foundation model for text-guided video editing. Released in September 2025, this model enables users to modify videos directly through natural language instructions, supporting a wide range of editing capabilities without requiring manual masks or annotations.

Core Technical Capabilities

Lucy Edit Dev introduces several powerful features that set it apart from existing video editing solutions:

Direct instruction-guided video editing using free-text prompts
Motion preservation across complex transformations
Identity conservation during character or clothing changes
Precise localized editing without collateral alterations
No requirement for manual segmentation or additional control signals
Supports 81-frame video generations with enhanced temporal consistency

Editing Versatility

The model demonstrates remarkable flexibility across multiple editing domains:

Clothing and accessory modifications
Character replacements
Object insertions and removals
Scene and environment transformations
Color and style adjustments

Technical Architecture

Built on the Wan2.2 5B architecture, Lucy Edit Dev leverages a sophisticated approach combining:

High-compression Variational Autoencoder (VAE)
Diffusion Transformer (DiT) stack
Rectified-flow generation mechanism
Channel concatenation for precise editing

'Our goal is to edit an input video V according to a natural-language instruction T, producing an output that reflects the requested edit while preserving identity, motion, and unedited regions,' the research team explained in their technical paper.