{"id":"02ec89fc-1529-4d2d-a309-997c7195d7a7","kind":"products","code":"UB-D7333","name":"Bernini","nameBasis":"original-brand","summary":"Bernini is an AI-native unified video generation and editing underlying architecture introduced by the ByteDance Bernini team. The product focuses on solving the practical pain points of the traditional video generation model in processing complex instructions, multi-reference object fusion, and long-line video editing, such as logic confusion, image incoherence, and dynamic control difficulties. [Product Features] Semantic programming engine based on multimodal large language model: The system automatically disassembles text instructions and reference materials at the bottom, and completes in-depth reasoning on spatiotemporal transformation and logical sequence before generation to ensure the coherence and accuracy of video logic. Multi-reference-guided video generation and editing: Supports the simultaneous input of up to five reference images to achieve accurate cloning and migration of characters' clothing, texture, ambient light and shadow, and specific styles. Seamlessly integrate the unified workflow of a variety of video editing capabilities: built-in advanced editing algorithms covering object addition and deletion, style conversion, weather simulation, focus shift, and expression control, etc., to meet the full link needs of film and television in the later stage. High stability and continuity of video generation rendering base: Advanced DiT rendering architecture and segmentation-aware position coding are used to ensure the structural stability of video under ultra-long camera tracks and complex motion sequences.","description":"Bernini is an AI-native unified video generation and editing underlying architecture introduced by the ByteDance Bernini team. The product focuses on solving the practical pain points of the traditional video generation model in processing complex instructions, multi-reference object fusion, and long-line video editing, such as logic confusion, image incoherence, and dynamic control difficulties. [Product Features] Semantic programming engine based on multimodal large language model: The system automatically disassembles text instructions and reference materials at the bottom, and completes in-depth reasoning on spatiotemporal transformation and logical sequence before generation to ensure the coherence and accuracy of video logic. Multi-reference-guided video generation and editing: Supports the simultaneous input of up to five reference images to achieve accurate cloning and migration of characters' clothing, texture, ambient light and shadow, and specific styles. Seamlessly integrate the unified workflow of a variety of video editing capabilities: built-in advanced editing algorithms covering object addition and deletion, style conversion, weather simulation, focus shift, and expression control, etc., to meet the full link needs of film and television in the later stage. High stability and continuity of video generation rendering base: Advanced DiT rendering architecture and segmentation-aware position coding are used to ensure the structural stability of video under ultra-long camera tracks and complex motion sequences.","country":null,"category":null,"originalName":"Bernini","eventDate":null,"website":null,"sourceUpdatedAt":"2026-10-31T03:20:18.435Z","englishUpdatedAt":"2026-11-01T00:48:57.975Z","reviewedAt":"2026-11-01T00:48:57.975Z","snapshotAt":"2026-11-01T00:41:14.697Z","sourceHash":"62ac0d21f4fad1542c72d96697ef1a256caaadc91673f7bcde0e92e1dbc952ca","companyId":"a1cbd485-2ab5-4673-9b5e-c10af5416eaa","foundedYear":null,"logoUrl":null,"originalText":"Bernini 是一款由字节跳动 Bernini 团队推出的 AI 原生统一视频生成与编辑底层架构。 该产品专注于解决传统视频生成模型在处理复杂指令、多参考物融合以及长线视频编辑时存在的逻辑混乱、画面不连贯及动态控制困难的实际痛点。【产品功能】基于多模态大语言模型的语义规划引擎： 系统在底层自动拆解文本指令与参考素材，在生成前完成对时空变换与逻辑顺序的深度推理，确保视频逻辑的连贯与准确。融合多参照物引导的视频生成与编辑： 支持同时输入多达五个参考图像，实现对人物服饰、材质质感、环境光影及特定风格的精准克隆与迁移。无缝融合多种视频编辑能力的统一工作流： 内置涵盖对象增删、风格转换、天气模拟、焦点转移及表情控制等多种高级编辑算法，满足影视后期全链路需求。高稳定性与连续性的视频生成渲染底座： 采用先进的 DiT 渲染架构与分段感知位置编码，确保视频在超长镜头轨道及复杂运动序列下的结构稳定性。","sources":[],"related":[{"id":"a1cbd485-2ab5-4673-9b5e-c10af5416eaa","kind":"companies","code":"UB-C13800","name":"Bernini","nameBasis":"source-english","summary":"Latent Semantic Planning for Video Diffusion","description":"Bernini is a unified framework for video generation and editing with self-supervised vision-text reasoning.","country":null,"category":null,"originalName":"Bernini","eventDate":null,"website":"https://bernini-ai.github.io/","sourceUpdatedAt":"2026-10-31T00:00:00.000Z","englishUpdatedAt":"2026-11-01T02:03:53.834Z","reviewedAt":"2026-11-01T02:03:53.834Z","snapshotAt":"2026-11-01T00:40:31.077Z","sourceHash":"e76514997984d3481fdd00f2bc95783fa14ab4af3f0d79ebd9d6d6d4118f5e9c","companyId":null,"foundedYear":null,"logoUrl":null,"path":"/companies/bernini"}],"path":"/products/bernini","structured":{"city":null,"country":null,"location":null,"role":null,"category":null,"company":{"name":"Bernini","path":"/companies/bernini"}},"definition":"Bernini is a product from Bernini.","summaryBasis":"translated","thin":false,"canonical":"https://uniquebase.ai/products/bernini"}