OpenAI has been revealed to be developing multiple deep updates for the ChatGPT web version, aiming to further reduce the "mechanical feel" of AI-generated content through a highly customized tool ecosystem.

According to a test interface leaked by developer Tibor Blaho, the core upgrade focuses on a new "writing template" feature. This feature allows users to upload personal email histories, articles, or specific documents as training samples, enabling the system to automatically extract and generate style templates that match the user's specific tone, structure, and expression habits. This marks a shift in generative AI from general "alignment with human preferences" to more private "alignment with individual characteristics," effectively solving the repeated prompt engineering issues when large models handle specific business or creative writing due to tonal misalignment.

QQ20260305-150111.png

In addition to deep adaptation of text style, OpenAI also plans to achieve breakthroughs in multimodal integration. Leaked information shows an added "animation" button that allows static images generated to be directly imported into video generation tools and automatically pre-fills motion prompts based on the image content, achieving a seamless transition from image to video. Furthermore, for professional office scenarios, ChatGPT may introduce an email writing mode that includes fields such as sender and cc, and supports exporting drafts directly in Markdown format.

This series of feature matrix exposure reflects that OpenAI is trying to enhance the product's productivity tool attributes by using refined function packages (Templates) and workflow integrations. In the current context of intensified homogenization competition among large models, improving the ability of AI to replicate user-specific styles not only enhances user stickiness but will also drive the industry to shift from a single model capability competition to a deeper user experience ecosystem construction.