商汤发布 SenseNova 6.1 CodeVideo 开源代码视频生成模型与 Skill
商汤发布 SenseNova 6.1 CodeVideo，一款可生成可执行代码来创作并渲染短视频的开源代码视频模型。该模型支持将创意转为视频并通过后续请求迭代，也能以可编辑代码复现和混编参考视频；官方同时开源模型权重、Skill、提示词及全部展示视频的完整生成轨迹，每段视频均由单个提示词在一次运行中生成。
模型、Skill、提示词和生成轨迹均已开放，读者可据此了解代码驱动短视频生成的具体流程与早期效果。
🎬 𝗗𝗲𝘀𝗰𝗿𝗶𝗯𝗲 𝘆𝗼𝘂𝗿 𝗶𝗱𝗲𝗮. 𝗔𝗜 𝘄𝗿𝗶𝘁𝗲𝘀 𝘁𝗵𝗲 𝗰𝗼𝗱𝗲. 𝗧𝗵𝗲 𝗰𝗼𝗱𝗲 𝗿𝗲𝗻𝗱𝗲𝗿𝘀 𝘁𝗵𝗲 𝘃𝗶𝗱𝗲𝗼.
Introducing 𝗦𝗲𝗻𝘀𝗲𝗡𝗼𝘃𝗮 𝟲.𝟴 𝗖𝗼𝗱𝗲𝟮𝗩𝗶𝗱𝗲𝗼, a 𝗹𝗶𝗴𝗵𝘁𝘄𝗲𝗶𝗴𝗵𝘁, 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 𝗺𝗼𝗱𝗲𝗹 that generates executable code to create and render short videos, with promising early results in visual composition and motion design.
With the Skill, you can:
🔹 Turn an idea into a video and refine it through follow-up requests.
🔹 Recreate and remix reference videos in editable code.
We’ve open-sourced the 𝗺𝗼𝗱𝗲𝗹 𝘄𝗲𝗶𝗴𝗵𝘁𝘀 𝗮𝗻𝗱 𝗦𝗸𝗶𝗹𝗹, plus the prompts and complete generation trajectories behind every showcase video. Each was generated from a 𝘀𝗶𝗻𝗴𝗹𝗲 𝗽𝗿𝗼𝗺𝗽𝘁 𝗶𝗻 𝗼𝗻𝗲 𝗿𝘂𝗻, using the Skill.
We see code as a bridge between ideas and finished work, with future research exploring video editing, animation and 3D creation.
🤗 Model: huggingface.co/sensenova/Sen…
💻 Skill & Showcases: github.com/OpenSenseNova/Sen…
来源：SenseTime_AI · x.com