Let’s look at the conclusion first

Click "Import Document" on the document editing page, and you can choose TXT, Markdown, Word or image documents. Files are read through the system document picker, and images are read through the system photo picker. Picture text is recognized natively by the Chinese OCR model built into the application, and supports Chinese, English, numbers, and mixed Chinese and English. Neither link requires additional application for album or storage permissions. After confirming the preview, you can insert the content under the current cursor, append it to the end of the text, or replace the entire text.

Operation steps

1. Create or open an editable document and click "Import Document" at the top of the page.
2. Select "TXT Text Document", "Markdown Document", "Word Document" or "Image Document" depending on the source. Word documents need to be .docx.
3. When importing files, select a file in the system document selector; when importing images, select 1 to 10 JPG, PNG, WebP or HEIC images in the system photo selector.
4. When importing multiple images, first drag to adjust the reading order, and then start recognition. It can be canceled during the recognition process; when a single piece fails, you can retry, explicitly skip the piece, or cancel the entire batch.
5. Check out editable previews and tips. Image documents can be switched between "Intelligent Segmentation" and "Keep Original Lines", and modifications in the two modes will be retained respectively.
6. Select "Insert at the current cursor", "Append to the end of the text" or "Replace the entire text".
7. Return to the edit page to check the title and text. Empty titles or "untitled documents" will take precedence over the file name and custom titles will not be overwritten; the results are then automatically saved.

Three ways to write

• Insert to the current cursor - Place the imported content at the text cursor, suitable for adding a paragraph or chapter.
• Append to end of text: Keep the existing text and append the entire file content to the end.
• Replace all text: Replace the current text with the imported file. Valid hotword mappings that are no longer in the new text may be removed, so check the preview first. Inserting and appending do not modify hotwords.

Which formats will be retained

• TXT - Preserves paragraphs and line breaks, supports UTF-8, UTF-16 LE/BE and GB18030; tabs are converted to spaces.
• Markdown - retain titles, boldface, lists, quotes, code text, link text, and table text; italics are treated as normal text, and images and HTML tags are skipped.
• DOCX - retain paragraphs, titles, boldface, underlining, numbering or item lists, table text; pictures, formulas, text boxes, comments, footnotes, headers and footers, etc. will be skipped.
• Image scripts: Recognize JPG, PNG, static WebP, and device-decodable HEIC/HEIF. The app handles orientation, long-image slicing, and overlap removal, then restores natural sections for single-column scripts where possible. Columns, tables, handwriting, decorative text, reflections, and severe blur may reduce accuracy.

Importing for easy reading and continued editing will not copy the Word page layout. When format degradation occurs, the preview will focus on the description and confirm that the main text is still available before writing.

File limits and failure handling

• TXT and Markdown files can be up to 10 MiB, and DOCX files can be up to 50 MiB; the converted text can be up to about 200,000 characters.
• There can be a maximum of 10 pictures at a time, with a maximum size of 30 MiB for a single picture and a maximum size of 120 MiB for a whole batch; a single picture is also limited to 60 million pixels, 12000px wide and 30000px high. The converted text is also up to about 200,000 characters.
• Legacy .docs, PDFs, password-protected DOCX, macro documents, GIFs, dynamic WebP, or cloud URLs are not supported.
• If the prompt format does not match, the file is damaged, the file is too large, the reading fails, or the parsing times out, the content will not be written into the text. Please go back and confirm whether the extension and file can be opened normally, and then select again.
• If pictures, formulas, etc. are skipped, it is a prompt that the import can continue; the import will be terminated when the text is empty or the file cannot be safely parsed.
• OCR results may incorrectly identify similar words, punctuation, English spaces, or paragraph boundaries. When you see "low confidence", "complex layout" or "long image has been split", you should check it against the original image in the preview before importing.