Bernini AI: Generate & Edit Video with ByteDance's Unified Model

Bernini AI is a cutting-edge platform that leverages ByteDance's open-source Bernini framework for AI-powered video generation and editing. This unified model allows users to create video content from text prompts, restyle existing footage using simple text instructions, and generate new clips based on reference images. It is built upon a sophisticated architecture that pairs a multimodal large language model (MLLM) based semantic planner with a Diffusion Transformer (DiT) based renderer.

Key Features:

• Text-to-Video Generation: Create video content from descriptive text prompts.
• Reference-to-Video: Generate new video clips by providing up to five reference images to anchor the subject, object, or style.
• Prompt-Based Video Editing: Modify existing footage by describing desired changes, such as restyling clips, swapping objects, or altering environmental conditions like weather.
• Unified Workflow: Bernini AI integrates generation and editing tasks within a single interface, eliminating the need for separate tools.
• Advanced Architecture: Powered by a Qwen2.5-VL based MLLM planner for semantic understanding and a Wan2.2 based DiT renderer for pixel-level synthesis.
• Accessibility: Usable directly in the browser without requiring multi-GPU setups or local installations, removing the barrier of high-end hardware like H100 GPUs.
• Open Source Foundation: Built on Bernini, an open-source model released under the Apache 2.0 license, allowing for inspection and community contribution.
• Editing Consistency: Excels in maintaining subject identity and stability in unedited regions of video clips, making it ideal for targeted edits.
• Additional Engines: Access to other AI video and image engines like Kling, Veo, GPT Image, and Seedream within the same workspace for diverse creative needs.

Bernini AI offers a powerful yet accessible solution for creators looking to generate and edit high-quality video content efficiently, leveraging state-of-the-art AI technology from ByteDance.