怎么样用AI写作助手完成Sora视频创作
ChatGPT发布仅过去了一年，市面上<span style="color: black;">已然</span><span style="color: black;">显现</span>了众多的AI写作辅助工具。而此时我觉得<span style="color: black;">选取</span>一款免费的AI写作助手，最重要的影响<span style="color: black;">原因</span>是<strong style="color: blue;">能否满足<span style="color: black;">各样</span>新型应用场景的需求</strong>，而不只限于<span style="color: black;">平常</span>的写邮件写文案等任务。而<span style="color: black;">近期</span>AI界突破性的<span style="color: black;">发展</span>当属Sora生成视频的发布，其生成的视频在流畅性，连贯性，<span style="color: black;">恰当</span>性，<span style="color: black;">连续</span>时间等方面都是超越此前已有的任何视频生成工具。<span style="color: black;">因此</span>我的近期写作应用场景，<span style="color: black;">便是</span>围绕着AI视频创作展开。一个AI视频创造<span style="color: black;">能够</span>分为<span style="color: black;">重点</span>三大<span style="color: black;">过程</span>：<strong style="color: blue;"><span style="color: black;">阅读理解剧本</span></strong>，<span style="color: black;"><strong style="color: blue;">分镜脚本管理</strong></span>，和<span style="color: black;"><strong style="color: blue;"><span style="color: black;">最后</span>的画面生成</strong></span>。而<span style="color: black;">怎样</span>去描述一个个电影级的镜头画面，<span style="color: black;">构成</span>一个连贯<span style="color: black;">恰当</span>的故事短片，<span style="color: black;">有些</span>普通的AI写作工具是难以完成的。<span style="color: black;">由于</span>这个<span style="color: black;">行业</span>比较垂直，缺乏语料训练。<span style="color: black;">因此</span>就<span style="color: black;">必须</span><span style="color: black;">这般</span>的AI写作助手，它<span style="color: black;">必须</span>能够处理分析整部电影剧本，<span style="color: black;">而后</span><span style="color: black;">经过</span>理解学习模仿，<span style="color: black;">拥有</span>进一步再创作的能力。<h2 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">阅读理解剧本</span></h2><span style="color: black;">因此</span>在调研试用了许多AI写作工具之后，我对Kimi.ai的表现印象深刻。我<span style="color: black;">重点</span>看重的是它<strong style="color: blue;"><span style="color: black;">拥有</span>20万字超长文本输入的阅读能力</strong>，<span style="color: black;">亦</span><span style="color: black;">便是</span>直接上传<span style="color: black;">全部</span>剧本进行阅读。<span style="color: black;">况且</span>是免费的，国内直接<span style="color: black;">能够</span>登录。<span style="color: black;">这儿</span>我<span style="color: black;">选取</span>了著名的科幻电影《阿凡达》的英文剧本进行测试，总共有157页，<span style="color: black;">能够</span>直接上传对任意场景分析无压力。我对这部电影印象最深刻的一个场景，<span style="color: black;">亦</span>是全剧最高潮的一段，<span style="color: black;">便是</span>主人公骑着巨大红色的飞龙，降临在阿凡达族群中，给当时处在绝境的人们带来了<span style="color: black;">期盼</span>。<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_jpg/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJM5BYiaZTnX8eIbicicpK8V3H3sXjVMEGbnTibTCFO2j5oxTLw2hQug6nZVw/640?wx_fmt=jpeg&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p><span style="color: black;">然则</span>时隔这么<span style="color: black;">数年</span>，主人公叫什么名字，那个红色龙叫什么，还有<span style="color: black;">她们</span>聚集在<span style="color: black;">一块</span>祈祷的<span style="color: black;">地区</span>我<span style="color: black;">皆想</span>不起来了，只能尝试的用模糊的描述<span style="color: black;">瞧瞧</span>Kimi能否理解我的需求<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMNjqk46twzdQiaAtvOvW1rB30Cq9GCrVTkic1hHx4XvBdCl9ib9aibX95Vg/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p><p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJM1j9nTEv2fzZAY8cQsHkDBq34ySHPLFdTmniaEYsmVHAMNq4rlKbU9Gw/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>说实话我<span style="color: black;">第1</span>次看见回复时候还是很惊艳的，之前<span style="color: black;">亦</span>尝试过<span style="color: black;">有些</span>其他AI<span style="color: black;">制品</span>类似的检索<span style="color: black;">加强</span>生成（RAG）文本的功能，但<span style="color: black;">常常</span>只能是胡乱截取片段回复的，缺乏<span style="color: black;">关联</span>性。而Kimi能够准确理解我模糊表达的含义，并且在<span style="color: black;">全部</span>剧本中准确检索到原文信息，还是非常厉害的。<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMNj5mW6JJr3f3nT8Ugm3ib5icvsg6R13ic65PdzsicEc3P7JxpkkWBaz5ibA/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">阿凡达原始剧本中对这一幕的描述和我要找的一模<span style="color: black;">同样</span></span></p>其实这<span style="color: black;">亦</span>是<span style="color: black;">日前</span>大模型<span style="color: black;">开发</span>中最<span style="color: black;">必须</span>改进的关键部分，<span style="color: black;">便是</span><span style="color: black;">怎样</span>减少幻觉生成，并且<span style="color: black;">加强</span>超长上下文语境理解能力。Kimi的<span style="color: black;">开发</span>团队<span style="color: black;">亦</span>对这种“大海捞针”的长文本性能进行了完整测试，在对比实验中，<strong style="color: blue;">Kimi的性能都超过了GPT-4 turbo以及Claude 2.1等其他模型</strong>，感兴趣的<span style="color: black;">伴侣</span><span style="color: black;">能够</span>阅读<span style="color: black;">她们</span>的完整报告。<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><a style="color: black;">Kimi Chat <span style="color: black;">颁布</span>“大海捞针”长文本压测结果，<span style="color: black;">亦</span>搞清楚了这项测试的精髓</a></p>
<h2 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">分镜脚本管理</span></h2><span style="color: black;">那样</span>理解电影剧本只是<span style="color: black;">第1</span>步，<span style="color: black;">怎样</span>学习模仿并再创造，<span style="color: black;">首要</span>就<span style="color: black;">必须</span>理解电影镜头语言。<span style="color: black;">实质</span>上，有一系列的术语描述<span style="color: black;">各样</span>镜头种类。我<span style="color: black;">亦</span>让Kimi对《阿凡达》剧本中的镜头画面进行翻译和解释。<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMJzdbCxDSHeDeISwJdbba2sbgLX0f8TQZ5xic5QibmncCVbSTUT1Q1sQQ/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>而在<span style="color: black;">认识</span>了这些电影术语之后，<span style="color: black;">咱们</span>就<span style="color: black;">能够</span>尝试去拆解这部剧本。其中一个重要环节<span style="color: black;">便是</span>分镜（storyboard），<span style="color: black;">亦</span><span style="color: black;">便是</span>以故事图像的可视化方式<span style="color: black;">来讲</span>明影片的<span style="color: black;">形成</span>，<span style="color: black;">通常</span>以一次运镜<span style="color: black;">做为</span>分解单位，并标注上镜头类型、时长、对白等<span style="color: black;">仔细</span>信息。例如，<span style="color: black;">咱们</span>可以让Kimi<span style="color: black;">按照</span>《阿凡达》剧本创建一个分镜管理表<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMAxqTNpW9iajjFoXYJRH8NcJb3q6rJ0ib3Pk3nzQ5dNyzSdbD9GDb5pCQ/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<h2 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;"><span style="color: black;">最后</span>画面生成</span></h2><span style="color: black;">因此</span>有了<span style="color: black;">这般</span>的分镜表，就<span style="color: black;">能够</span>一方面<span style="color: black;">按照</span>画面内容描述生成影片，<span style="color: black;">同期</span>镜头类型、运镜方式以及对白配音等其他元素<span style="color: black;">亦</span>进一步丰富影片的变化细节。那在AIGC时代，一切都<span style="color: black;">能够</span>用AI生成。Sora模型<span style="color: black;">能够</span>用<span style="color: black;">照片</span>输入<span style="color: black;">做为</span><span style="color: black;">初始</span>，生成完整影片。<span style="color: black;">那样</span><span style="color: black;">咱们</span><span style="color: black;">亦</span><span style="color: black;">能够</span>用Kimi来辅助生成描述<span style="color: black;">繁杂</span>电影画面的提示词，再<span style="color: black;">运用</span>类似Midjourney或Stable Diffusion等文生图工具进行绘制。<span style="color: black;">这儿</span>我让Kimi模仿《阿凡达》剧本格式进行再创作，生成一部<strong style="color: blue;">历史冒险类的短片</strong>，主题内容和镜头类型都由Kimi决定，并且输出相应的文本生成<span style="color: black;">照片</span>的prompt<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMv9M0DRhhzQiaG81G7d8cJRRiaqbia7s4CibuyeC8KbEQJFdibhBS7vKQTxQ/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>我<span style="color: black;">按照</span>上面生成prompt生<span style="color: black;">成为了</span>分镜脚本板，看起来还不错，<span style="color: black;">已然</span>有一点故事感了<img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SxffuSM3T7ialD4tVg4jrqcic7ZhNCia1vQhjdMcrIsZo0Lw7vKJbG3J4hhs2sUBK2G1Coia6rULBHThQ/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"><span style="color: black;">dawn breaks over an ancient clock tower shrouded in mist, cinematic, oil painting style, consistent with a historical drama theme；黎明破晓于一座被雾气<span style="color: black;">包围</span>的古老钟楼上，<span style="color: black;">拥有</span>电影感，油画风格，与历史剧主题相一致。</span>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SxffuSM3T7ialD4tVg4jrqcicnb1zicbh0hcjIcdials1icZOCSV10hsEFuskSIAEKt9UsxFAGGglAhjicA/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"><span style="color: black;">close-up of an elderly clockmakers hands delicately repairing an antique clock, detailed, with a focus on craftsmanship, consistent with a historical drama theme；</span><span style="color: black;">特写镜头展示一位年迈钟表匠的双手正在精细地修复一座古董钟表，注重细节，强调工艺，与历史剧主题相符。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SxffuSM3T7ialD4tVg4jrqcicwRzuRJ9lsiblf97vbbDyF58jNINXeFVOFdXoMA45iaPIM2fmOdxJV5UA/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"><span style="color: black;">a close-up shot of a gear slowly turning in the hands of a skilled clockmaker, high detail, reminiscent of a classic timepiece, consistent with a historical drama theme；</span><span style="color: black;">一个特写镜头<span style="color: black;">捉捕</span>到一位技艺高超的钟表匠手中缓缓转动的齿轮，细节丰富，让人联想到经典的时计，与历史剧主题保持一致。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMIMeCgmhrDcTWkmsGFmX3pohgrwbDY5AWmbuCMM6AAnKZBSBL3qtEdw/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">a panoramic view of a small town waking up to the sound of the town clock, with people starting their day, consistent with a historical drama theme；</span><span style="color: black;">一个全景视角展示小镇随着镇钟的声响苏醒，人们<span style="color: black;">起始</span><span style="color: black;">她们</span>的一天，与历史剧主题相符合。</span></span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMQ2yFk28ULlLfU8d7zic4JuLaDVlbDpkZrnvBfH7ck5zgrwm7uqu5uDA/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">a tracking shot following a young woman walking through a quaint town with a yello</span><span style="color: black;">wed letter in hand, searching for an address, consistent with a historical drama theme；</span><span style="color: black;">一个跟随镜头，<span style="color: black;">捉捕</span>一位<span style="color: black;">青年</span>女子手持一封泛黄的信件，穿行在古色古香的小镇上，寻找一个<span style="color: black;">位置</span>，这一场景与历史剧主题相契</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/gayU3F5s3SzXMSDpaDQbS2HRz7w5WZJMbeTst6icjynH7ibfuYwg0kxUTAXrdXUp3zCqBIqgic5kWbbibF06tws9Xg/640?wx_fmt=png&from=appmsg&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">a dialogue scene between a young woman and an elderly clockmaker discussing the legend of "the rift in time," consistent with a historical drama theme；</span><span style="color: black;">一个对</span><span style="color: black;">话场景，一位<span style="color: black;">青年</span>女子与一位年迈的钟表匠讨论“时间裂缝”的传说，这一主题与历史剧的风格保持一致。</span></span></p>
<h2 style="color: black; text-align: left; margin-bottom: 10px;"><span style="color: black;">结语</span></h2>总结下来Kimi在长文本视频剧本创作方面的能力非常出色，这<span style="color: black;">亦</span>与其<span style="color: black;">背面</span>优秀的技术团队——月之暗面（Moonshot AI）的实力密切<span style="color: black;">关联</span>。<strong style="color: blue;">其创始人杨植麟<span style="color: black;">亦</span>是清华计算机系有名的AI专家，<span style="color: black;">熟练</span><span style="color: black;">认识</span>自然语言处理<span style="color: black;">行业</span>的都<span style="color: black;">晓得</span>他的<span style="color: black;">表率</span>作XLNet以及Transformer-XL模型</strong>。而随着Sora模型的推出<span style="color: black;">导致</span>了AI发展的新热潮，<span style="color: black;">将来</span>结合像Kimi<span style="color: black;">这般</span>的文本对话助手和Sora<span style="color: black;">这般</span>的视频图像处理模型，简单的文字将会赋予<span style="color: black;">每一个</span>人无限丰富的自我表达能力和表现形式，<span style="color: black;">将来</span>充满无限可能令人期待。官网链接是http://kimi.ai，<span style="color: black;">大众</span><span style="color: black;">亦</span>可<span style="color: black;">以避免</span>费尝试Kimi写作助手 请问、你好、求解、谁知道等。 太棒了、厉害、为你打call、点赞、非常精彩等。 论坛外链网http://www.fok120.com/ 哈哈、笑死我了、太搞笑了吧等。 你的留言真是温暖如春，让我感受到了无尽的支持与鼓励。
页:
[1]