Skip to main content

Style and format learning

Overview

AI Scribe's style and format learning feature analyzes your documents to create a custom style model. You can then use this model to generate drafts that match your preferred writing style and format. This feature lets you train models without code, simplifying the process.

Access the feature

Go to the AI Scribe workbench and click write from draft.
2

Procedure

Step 1: Create a style model

  1. On the write from draft page, upload local assets or add articles from the asset library.
    • Click upload local assets to upload files from your local machine.
    • Click add from asset library. You can search for and add assets from the asset library, or add custom text.
      image
    The following figure shows the page after you add documents:
    image
    • You must upload at least one document, and the total character count must not exceed 50,000.
    • The model cannot learn from documents that are niche, highly formulaic, or use rigid formatting like tables. You must provide a description of the stylistic and formatting standards for any documents you upload.
  2. Click start analysis to analyze the source documents. The following figure shows the page after the analysis is complete.
    During document analysis, you can Terminate Generation.
    3
    • Click Optimize Again to re-analyze and summarize the uploaded documents.
    • Click manual adjustment to edit the content generated by the model.
      4
    • Click discard to discard all initial analysis results.
  3. Once satisfied with the analysis results, click accept and name style. Enter a name to create the style.
    5

Step 2: Use the style

  1. Click use style to create a draft.
    7
  2. The following figure shows the create draft page. You can select your desired style from the style tab.
    8

Additional features

Refine an existing style

Click enhance to refine an existing style by re-editing and analyzing its source documents.
10

Provide feedback

Click submit feedback to provide expected results and suggestions for improvement on the current style model.
9