行业

Adobe 称通用模型在创意精确度上仍有不足

精选理由

这是 Adobe 在自家 MAX 2026 大会上的说法,可以和各家通用模型的营销口径对照着看。做企业级内容生产的团队值得留意:角色一致性、帧间持久上下文这些坑,很多人用通用模型时都踩过。

在 Adobe MAX 2026 大会上,Adobe 的 Hannah Elsakr 表示,通用 AI 模型在创意工作流中的精确度控制上仍无法满足企业需求。

正文

她指出,多数企业高管认为直接使用现成(off-the-shelf)模型不够,即便通过上下文工程(context engineering)提供参考素材,也只能前进一小步。企业真正需要的是跨时间的持久上下文:从一帧到另一帧保持角色外观一致,同时保留创作所需的知识、动作与语境。

这段发言以 Adobe MAX 2026 为背景,指向企业在搭建 AI 创意管线(creative pipelines)时的实际痛点。

原文 · The Information

General-purpose AI models are getting better. But Adobe says they still fall short when it comes to creative precision.

On the Ground at Adobe MAX 2026, Adobe's Hannah Elsakr explains why enterprises need more than generic models if they want to build AI-enabled creative pipelines.

"When it comes to the precision of artistry, most C-suite execs will tell you that off-the-shelf isn't enough. With context engineering, you can get a little further, providing references. But persistent context over time is what is the need at an enterprise. We can't forget what this character looks like from frame to frame…making sure you preserve the knowledge, the motion, the context for what it means to create. Off-the-shelf just can’t do that."

Sponsored by @Adobe.

产业坐标与信源核验

本条情报归属 AITOP「行业」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。涉及核心实体与技术议题包括:Adobe、context engineering、创意工作流、AI视频生成,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。

信源核验说明:本文原始线索采编自「The Information」(@theinformation),经 AITOP 多维评分管线结构化提炼。