Better documentation for --print-parameters.

[idrise]

Your Feature Request

The documentation for each parameter is super short and hard for an end user to understand. It would be good to understand which parameters apply to the new LSTM and which are the legacy OCR.

Changing --print-parameters might be a breaking change but adding a manual file would be harmless and useful.

[zeel2104]

I would like to work on this. Can you assign it to me?

[stweil]

Do you want to use a chatbot to do this task? It might be a real challenge to create correct documentation for each parameter.

[zeel2104]

Thank you for the suggestion! I understand that documenting all the parameters correctly could be challenging. I can explore using a chatbot to help generate clear explanations for each parameter and then review to ensure accuracy.

I can start working on it and provide an update once I have a draft.

[idrise]

If it started with the most well known and modern parameters first with great documentation like for instance this was generated by Gemini for the top ten most used in it's humble opinion. I mean it's not perfect, char_blacklist also had availability issues. As it gets to more obscure parameters it might need to analyse the source code using something like claude code or codex to try and uncover what..

chop_split_dist_knob 0.5 Split length adjustment

does. 😕

1. tessedit_char_whitelist

• Default: "" (Empty)
• What it does: Forces Tesseract to only recognize the characters you provide in the string.
• When to use it: When you know the exact character set of your data. For example, if you are reading a gas meter, zip codes, or phone numbers, you can set -c tessedit_char_whitelist=0123456789. This stops Tesseract from mistaking a "0" for an "O" or a "1" for an "I". (Note: In Tesseract 4, this requires version 4.1+ to work properly with the LSTM engine).

2. tessedit_char_blacklist

• Default: "" (Empty)
• What it does: The exact opposite of the whitelist. It forbids Tesseract from outputting specific characters.
• When to use it: If Tesseract consistently hallucinates special characters (like |, ~, or #) in a document that you know only contains standard alphanumeric text, you can blacklist those specific symbols.

3. preserve_interword_spaces

• Default: 0 (False)
• What it does: Instructs the engine to keep multiple spaces between words instead of collapsing them into a single space.
• When to use it: This is incredibly useful when OCR-ing tabular data, receipts, or text with monospace formatting. It helps preserve the visual structure and layout of columns in plain text output.

4. load_system_dawg

• Default: 1 (True)
• What it does: Toggles Tesseract's built-in system dictionary (Directed Acyclic Word Graph).
• When to use it: Tesseract naturally tries to force character combinations into known dictionary words. If you are trying to OCR random alphanumeric strings, URLs, passwords, or serial numbers, Tesseract might "correct" them into real words. Passing -c load_system_dawg=0 stops this behavior.

5. load_freq_dawg

• Default: 1 (True)
• What it does: Toggles Tesseract's dictionary for the most frequent words.
• When to use it: Almost always used in conjunction with disabling load_system_dawg. By passing -c load_freq_dawg=0, you completely free Tesseract from linguistic constraints, forcing it to rely entirely on character shape recognition rather than word probabilities.

6. tessedit_write_images

• Default: 0 (False)
• What it does: Saves Tesseract's internal, pre-processed image to your working directory (usually named tessinput.tif).
• When to use it: This is the ultimate debugging tool. Before Tesseract reads text, it converts your image to black-and-white (binarization/thresholding). If your OCR results are terrible, setting -c tessedit_write_images=1 allows you to see what the engine is actually "looking" at. If the output image is a blurry, noisy mess, you know you need to pre-process your image before feeding it to Tesseract.

7. user_words_suffix

• Default: "" (Empty)
• What it does: Allows you to supply a custom dictionary of words.
• When to use it: If you are scanning medical documents, legal text, or domain-specific jargon, Tesseract might struggle with the terminology. By providing a custom word list (e.g., -c user_words_suffix=user-words), Tesseract will actively look for and properly identify your custom terminology.

8. tessedit_create_tsv (and tessedit_create_hocr / tessedit_create_pdf)

• Default: 0 (False)
• What it does: Changes the standard plain-text output into different structured formats.
• When to use it: If you need to know exactly where a word is on a page or how confident Tesseract is in its prediction, setting -c tessedit_create_tsv=1 forces the output to be a Tab-Separated Values file containing X/Y bounding box coordinates and confidence scores (from 0 to 100) for every single recognized word.

9. textord_min_linesize

• Default: 1.25
• What it does: Determines the minimum height threshold Tesseract uses when analyzing the layout to detect lines of text.
• When to use it: If you are OCR-ing a document and noticing that Tesseract is randomly breaking a single line of text into multiple broken lines (or mashing lines together), bumping this value up (e.g., -c textord_min_linesize=3.0) can often fix the layout analysis.

10. tessedit_do_invert

• Default: 0 (False)
• What it does: Tells the engine to invert the colors of the image internally.
• When to use it: Tesseract natively assumes that documents consist of dark text on a light background. If you are scanning an image with white text on a black background, the engine may fail entirely. While newer versions attempt to auto-detect this, setting -c tessedit_do_invert=1 explicitly forces it to read inverted text.

[stweil]

Copilot created a first draft. Please add comments and suggest changes for pull request #4526.