mirror of
https://github.com/Ladebeze66/llm_ticket3.git
synced 2025-12-13 10:46:51 +01:00
101 lines
2.9 KiB
Plaintext
101 lines
2.9 KiB
Plaintext
ocr:
|
|
"""
|
|
Your task is to extract all visible text from the provided image with absolute completeness.
|
|
|
|
- Do not interpret, summarize, rephrase or explain the content.
|
|
- Do not add anything that is not explicitly written in the image.
|
|
- Do not describe the layout, style, or appearance.
|
|
- Do not translate. Return text in its original language.
|
|
|
|
Return every visible string, including those that are:
|
|
- Small, low contrast, partially hidden or truncated
|
|
- Located in headers, sidebars, buttons, labels, fields, tooltips, or footers
|
|
|
|
Segment your extraction into the following visual zones:
|
|
- Top
|
|
- Left
|
|
- Center
|
|
- Bottom-right
|
|
- Bottom-left
|
|
|
|
Under each section, return only the raw text seen in that area, preserving line order (top to bottom).
|
|
|
|
Format:
|
|
Top:
|
|
* text line
|
|
Left:
|
|
* text line
|
|
...
|
|
|
|
Only include a zone if at least one string is found in that area. Skip empty sections silently.
|
|
"""
|
|
|
|
"""
|
|
Your task is to extract all clearly visible or partially visible text from the image.
|
|
|
|
Segment the result into the following visual zones:
|
|
- Top
|
|
- Left
|
|
- Center
|
|
- Bottom-left
|
|
- Bottom-right
|
|
|
|
Rules:
|
|
- You may complete any partially cropped text if it is strongly implied.
|
|
- Avoid high-level summaries like "no text can be extracted".
|
|
- Do not describe dynamic UI behavior or HTML logic — only extract what's on screen.
|
|
- Return lines as-is, including those found inside buttons, menus, or input fields.
|
|
- No descriptions or commentary.
|
|
|
|
If you can read even a portion of a line, include it. Do not skip partial strings.
|
|
|
|
Format:
|
|
Top:
|
|
* text...
|
|
Left:
|
|
* text...
|
|
Center:
|
|
* text...
|
|
Bottom-left:
|
|
* text...
|
|
Bottom-right:
|
|
* text...
|
|
|
|
Do not include commentary like "no visible text". Simply omit empty zones.
|
|
"""
|
|
"""
|
|
Your task is to extract all visible text from the provided image with absolute completeness.
|
|
|
|
- Do not interpret, summarize, rephrase or explain the content.
|
|
- Do not add anything that is not explicitly written in the image.
|
|
- Do not describe the layout, style, or appearance.
|
|
- Do not translate. Return text in its original language.
|
|
|
|
Return every visible string, including those that are:
|
|
- Small, low contrast, partially hidden or truncated
|
|
- Located in headers, sidebars, buttons, labels, fields, tooltips, or footers
|
|
|
|
Segment your extraction into the following visual zones:
|
|
- Top (header or navigation bar)
|
|
- Left (sidebar or vertical panel)
|
|
- Center (main content)
|
|
- Bottom-right (lower-right corner)
|
|
- Bottom-left (lower-left corner)
|
|
|
|
Under each section, return only the raw text seen in that area, preserving line order (top to bottom).
|
|
|
|
Strict instructions:
|
|
- Do NOT skip lines that are cropped, faint or small.
|
|
- Do NOT say "No visible text". Omit a section completely if nothing is seen.
|
|
- Include interface elements (buttons, labels, menus, footnotes).
|
|
|
|
Format:
|
|
Top:
|
|
* text line
|
|
Left:
|
|
* text line
|
|
...
|
|
|
|
Only include a zone if at least one string is found in that area. Skip empty sections silently.
|
|
""
|