Lucy Edit Dev – Decart AI’s open-source text-instruction-based video editing model

What is Lucy Edit Dev？

Lucy Edit Dev is an open-source text-instruction-based video editing model developed by the Decart AI team. It can perform a variety of video editing operations based on simple text prompts, such as changing clothing, replacing characters, inserting objects, and transforming scenes—while perfectly preserving motion and composition in the video. The model is built on the Wan2.2 5B architecture, inheriting the high-compression VAE + DiT stack, making it easy for users to adapt existing scripts and workflows.

Key Features of Lucy Edit Dev

Text-instruction-based video editing: Users can guide video editing with plain text instructions, without fine-tuning or using masks—simple and intuitive.
Multiple editing types: Supports changes in clothing and accessories, character replacement, object insertion, scene transformation, and more to meet diverse editing needs.
Motion and composition preservation: Precisely retains the subject’s movements and the overall composition of the video, ensuring natural fluidity.
High-precision editing: Accurately preserves character identity and motion, ensuring consistency between the edited video and the original.
Open-source architecture: Built on the Wan2.2 5B framework with a high-compression VAE + DiT stack, making integration into existing workflows easier.

Technical Principles of Lucy Edit Dev

Text-driven editing: Uses natural language processing to parse text instructions and understand user editing intentions.
Deep learning architecture: Based on the Wan2.2 5B framework, leveraging a high-compression VAE (Variational Autoencoder) and DiT (Diffusion Transformer) stack for efficient video editing.
Video frame processing: Breaks down videos into individual frames, applies editing to each frame, while maintaining temporal coherence and consistency.
Motion and composition preservation: Employs advanced motion estimation and composition analysis to ensure naturalness and stability in edited results.
Text-to-video mapping: Maps textual instructions to video content, translating editing intentions into concrete video edits via generative modeling.
No fine-tuning required: Allows direct text-based editing without model fine-tuning or mask usage, simplifying the workflow.

Project Links

GitHub: https://github.com/DecartAI/lucy-edit-comfyui
HuggingFace: https://huggingface.co/decart-ai/Lucy-Edit-Dev

Application Scenarios of Lucy Edit Dev

Video content creation: Creators can quickly modify video elements, such as changing outfits or adding effects, to boost productivity.
Advertising production: Marketing teams can adjust product displays, settings, or scenes in ads to fit different audience segments and strategies.
Film and TV post-production: Enables operations like character replacement and scene transformation, reducing time and cost in post-production.
Animation production: Animators can easily modify character appearances or scenes through text prompts, accelerating the animation workflow.
Educational video creation: Educators can adapt teaching videos by adjusting visual elements to better meet instructional needs.
Social media content optimization: Users can tailor video content for different platforms and audience preferences, improving engagement rates.