Many-for-Many: Unify the Training of Multiple Video and Image Generation and Manipulation Tasks

2025-06-02

AI Summary

Current vision generative models are predominantly single-task architectures reliant on large-scale, high-quality annotated datasets, resulting in limited generalization and narrow task coverage. To address this, we propose the first many-for-many multi-task unified framework, trained end-to-end from scratch on heterogeneous multimodal data—including text-to-image/video, image-to-video, video-to-video, and diverse image/video editing tasks—using lightweight conditional adapters and joint image-video progressive training. A key innovation is the incorporation of depth maps as 3D spatial priors, enabling unified support for over ten distinct generation and editing tasks (e.g., T2V, I2V, V2V, inpainting, motion transfer). We release two open-source model variants: a high-fidelity 8B-parameter version whose video generation quality rivals leading open and commercial engines, and a compact 2B-parameter variant. All code and models are publicly available.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Diffusion models have shown impressive performance in many visual generation and manipulation tasks. Many existing methods focus on training a model for a specific task, especially, text-to-video (T2V) generation, while many other works focus on finetuning the pretrained T2V model for image-to-video (I2V), video-to-video (V2V), image and video manipulation tasks, etc. However, training a strong T2V foundation model requires a large amount of high-quality annotations, which is very costly. In addition, many existing models can perform only one or several tasks. In this work, we introduce a unified framework, namely many-for-many, which leverages the available training data from many different visual generation and manipulation tasks to train a single model for those different tasks. Specifically, we design a lightweight adapter to unify the different conditions in different tasks, then employ a joint image-video learning strategy to progressively train the model from scratch. Our joint learning leads to a unified visual generation and manipulation model with improved video generation performance. In addition, we introduce depth maps as a condition to help our model better perceive the 3D space in visual generation. Two versions of our model are trained with different model sizes (8B and 2B), each of which can perform more than 10 different tasks. In particular, our 8B model demonstrates highly competitive performance in video generation tasks compared to open-source and even commercial engines. Our models and source codes are available at https://github.com/leeruibin/MfM.git.

Problem

• Unified training for multiple video and image generation tasks
• Joint learning strategy improves video generation performance
• Lightweight adapter unifies conditions across diverse visual tasks

Innovation

• Unified framework for multiple visual tasks
• Lightweight adapter for different conditions
• Joint image-video learning strategy