Have you ever wanted to copy text from a scanned Santali book, an old magazine, or a photograph of an Ol Chiki document? Santali OCR can help you convert printed text into editable digital text without typing every character manually. This beginner-friendly guide explains what OCR is, why Santali needs better OCR tools, and how to use it on Windows, Android, and Linux.
1. What is Santali OCR?
OCR stands for Optical Character Recognition. It is a technology that reads text from images, photographs, scanned books, and image-based PDFs and converts it into digital text that you can copy, edit, search, and reuse.
For Santali, OCR is especially important for recognizing text written in Ol Chiki (ᱚᱞ ᱪᱤᱠᱤ). For example, a printed page containing Santali sentences can be scanned and converted into Unicode text that you can edit in LibreOffice Writer, Microsoft Word, or a website.
2. Why Does Santali Need OCR?
Many valuable Santali books, historical records, poems, stories, magazines, and educational materials are available only as printed pages or scanned images. Without OCR, readers often have to type the entire text manually before publishing or searching it online.
- Preserve literature: Digitize old Santali books and magazines.
- Support education: Prepare editable study materials for students and teachers.
- Build digital libraries: Make Santali documents searchable and easier to access.
- Support Wikimedia: Help prepare source text for proofreading and publication on Santali Wikisource.
- Develop language technology: Create datasets for dictionaries, translation systems, spell-checkers, and other language tools.
OCR does not replace proofreading. Instead, it reduces repetitive typing and gives Santali speakers a useful starting point for digitizing their language resources.
3. Understanding Santali Scripts
Santali is written in several scripts, depending on the region and publication. These include Ol Chiki, Bengali, Odia, Devanagari, and Roman script. Ol Chiki was created by Pandit Raghunath Murmu and is the dedicated script used for modern Santali writing.
| Script | Where you may encounter it |
|---|---|
| Ol Chiki | Modern Santali books, newspapers, and educational resources |
| Bengali | Some regional publications and historical documents |
| Odia | Some Santali publications in Odisha |
| Devanagari | Some publications and learning resources |
| Roman | Language-learning materials, linguistic documentation, and informal writing |
Important: An OCR model trained for one script cannot automatically recognize every other script. For Ol Chiki documents, select a model specifically intended for Santali in Ol Chiki.
4. Why Is Santali OCR Difficult?
English OCR has benefited from decades of development and large training datasets. Santali OCR has had fewer resources, making development more challenging.
- Limited training data: OCR systems need page images paired with correct text.
- Unique characters: Ol Chiki has its own character shapes, so Latin or Devanagari models are not a substitute for an appropriate Ol Chiki model.
- Old documents: Faded ink, damaged pages, unusual fonts, and uneven printing make recognition harder.
- Limited application support: Many general-purpose OCR applications do not provide a dedicated Santali model.
- Handwriting: Models trained mainly on printed books may struggle with handwritten notes.
This is why Google Lens or another general OCR application may return incorrect characters, omit marks, or produce meaningless text when processing Ol Chiki.
5. Are There Any Existing Models or Technologies for Santali Image-to-Text Conversion?
Yes, there have been several efforts to develop OCR (Optical Character Recognition) technology for Santali. However, reliable image-to-text conversion for Santali written in Ol Chiki remains an area that needs further development, testing, and community participation.
5.1 Indic OCR: An Earlier Effort
One earlier initiative was Indic OCR, which provided alternative Tesseract trained-data models for several Indian languages, including Santali. These models were trained using fonts such as Noto and were intended to improve OCR recognition for Indic scripts.
I discovered this project while exploring ways to convert Santali book scans into editable text. I initially tried using its existing Santali model and later attempted to work on OCR support myself, including experimenting with different fonts. However, the available model and training resources did not provide the modern recognition results I was looking for. I could not find a newer release that addressed my requirements.
My experience showed me that developing OCR for Santali is not simply a matter of installing software. We also need updated models, suitable training data, and testing on real Santali books printed in different fonts.
Explore the Indic OCR repository on GitHub to learn about its existing models and approach.
5.2 Government of India's Digital Text-Recognition Efforts
Another resource worth exploring is the Government of India's Adivaani portal. It is relevant to the wider effort to make tribal-language resources more accessible through digital technology.
Image-to-text functionality and related digital-language tools may provide additional opportunities for Santali users. However, I have not yet confirmed which OCR engine or underlying technology is used for the particular image-to-text service, or how well it recognizes Ol Chiki.
If I receive reliable information about the technology used, its supported scripts, or its Santali recognition accuracy, I will update this section. Until then, users should test the service with their own documents and verify the extracted text.
5.3 Sarvam Vision: An AI-Based OCR Technology
Sarvam AI has developed Sarvam Vision, a vision-language model designed for document digitization, OCR, and document understanding. Unlike traditional OCR systems that primarily recognize text, document-understanding systems can also process page structure, tables, and reading order.
I uploaded an image and it gave me detail analysis along with text extraction.
I'll zoom in on the text to read the Ol Chiki characters accurately.
The image is a single line of Santali written in the Ol Chiki script (ᱚᱞ ᱪᱤᱠᱤ). Here is the extracted text, character for character:
**ᱥᱟᱱᱟᱢ ᱠᱚ ᱡᱚᱦᱟᱨ ᱜᱮ ᱪᱮᱫ ᱞᱮᱠᱟ ᱢᱮᱱᱟᱜ ᱯᱮᱭᱟ**
Word by word:
| Ol Chiki | Romanised | Meaning |
|---|---|---|
| ᱥᱟᱱᱟᱢ | sānam | all / everyone |
| ᱠᱚ | ko | plural marker |
| ᱡᱚᱦᱟᱨ | johār | greetings (the traditional Santali greeting) |
| ᱜᱮ | ge | emphasis particle ("indeed / only") |
| ᱪᱮᱫ ᱞᱮᱠᱟ | ched lekhā | how / in what way |
| ᱢᱮᱱᱟᱜ | menāg | to be, to exist |
| ᱯᱮᱭᱟ | peyā | you (plural / respectful) |
So the whole line reads roughly: "Greetings to all — how are you all doing?"
According to the official Sarvam Vision documentation, the system supports 22 Indian languages and English, including Santali in its listed language set. Its document-processing capabilities can extract text from scanned documents and PDFs and return structured output such as HTML or Markdown.
However, a language appearing in a supported-language list does not automatically establish reliable recognition of every script used to write that language. Before using Sarvam Vision for Ol Chiki books, check its current language and script support, test representative pages, and proofread the results.
For more technical details, visit the Sarvam Vision article and the official documentation. Access methods, pricing, and usage limits may vary.
5.4 A New Santali LSTM Model by Subhashish
A promising recent development is the Santali LSTM model created by Subhashish Panigrahi. His contribution, titled Add Santali (sat) – first Ol Chiki script LSTM model (#15), was submitted to the Tesseract community in June 2026.
I also tried following the newer Tesseract 5.0 training approach to develop an updated model. However, my results were not satisfactory. Subhashish's work is therefore an encouraging development for people like me who want to improve Santali OCR.
The model uses an LSTM-based recognition approach and was trained using synthetic text and real scanned book pages. According to its initial report, recognition accuracy still needs improvement. Nevertheless, making an experimental model available gives the community a starting point for further testing and development.
I encourage Santali speakers, researchers, developers, and students to try this model and share their feedback in the comments below. Test it with different Ol Chiki fonts, old books, newspapers, and clean scanned pages. Please report what works, which characters are frequently misrecognized, and what improvements you would like to see.
Get the latest model and installation instructions from the Santali LSTM model contribution. Check the project's current status and instructions before downloading. The model is experimental, so proofread its output rather than assuming it is correct.
5. Which Technology Should Santali Users Try?
- For a quick online test: Try i2OCR Santali OCR.
- For local, open-source OCR: Try Tesseract with a compatible Santali model through gImageReader.
- For experimental Ol Chiki recognition: Explore Subhashish's Santali LSTM model.
- For AI-based document extraction: Investigate Sarvam Vision and verify its performance on Ol Chiki samples.
- For research and development: Explore the Indic OCR repository and the SantaliOCR project.
These options use different approaches, and their results should not be assumed to be equivalent. The most useful choice will depend on the document, font, image quality, and actual support for Ol Chiki.
Let's improve Santali OCR together. If you try any of these tools, please share your experience in the comments below. Tell us which model you used, what kind of document you tested, and whether the output was accurate. Community feedback, corrected text, and properly licensed training data can help build better OCR technology for Santali.
6. Method One: Use i2OCR Online
i2OCR Santali OCR is a convenient option for beginners because it works in a web browser without requiring a local Tesseract installation.
Steps to recognize Santali text
- Open the i2OCR Santali OCR webpage.
- Choose a clear photograph or scanned image containing printed Santali text in Ol Chiki.
- Select the Santali language option and review any available OCR engine or layout settings.
- Upload the image and start OCR.
- Select checkbox I'm not a robot.
- Review the extracted text in the result area.
- Copy the output or download it in an available format, such as TXT or DOCX.
- Open the output in a Unicode-compatible editor and proofread it against the original image.
For a scanned PDF, use the i2OCR Santali PDF OCR page. Follow the on-screen instructions to select the document and process the required page.
Advantages: It is convenient, requires no software installation, and is suitable for a quick first test.
Limitations: Free processing may be restricted, formatting may not be preserved, and recognition quality depends on the source image and the service's current OCR engine. Its internal model may differ from the community Tesseract LSTM model described above.
7. Method Two: Santali OCR on Windows Using gImageReader
gImageReader is a graphical application that provides an easier way to use the Tesseract OCR engine. Instead of entering every command in a terminal, you can open images or documents, select a recognition language, and view extracted text in a graphical interface. I have already created a video on gImageReader; please feel free to check it out below (uses old tessdata).
Step 1: Download the Santali model
Click on view GitHub and enter sat folder
Click on sat.traineddata
Click on the download raw file
Now transfer that file to a easy accessible location, ex. Desktop
Also, download the first LSTM model for Santali language created by OFDN foundation, Open the Santali LSTM model contribution (Experimental).
Download the sat.traineddata file the same way and save it.
Step 2: Install the software
Download a compatible Windows build of gImageReader from its official project page or linked releases. Check the bundled Tesseract version and install or extract the application according to its instructions.
- Go to the official website - https://gimagereader.com/
- Click on the download button
- I recommend downloading the latest release, as many bugs have been fixed. Head over to the gImageReader GitHub repository and check the releases section in the right sidebar.
- At the time of writing, the gImageReader version was 3.4.4.
- I recommend downloading the portable version since it doesn't require installation.
Step 3: Place the model in the tessdata folder
Locate the Tesseract data directory used by your gImageReader installation. Depending on the Windows build, it may be inside the application directory or a separate Tesseract installation. A common location for a bundled installation is:
C:\Program Files\gImageReader\share\tessdataThis is only an example; your actual directory may differ. If necessary, check gImageReader's settings or application information for the configured Tesseract data path. Copy sat.traineddata into the correct tessdata directory.
Step 4: Recognize the document
- Restart gImageReader, or use its language-redetection option if available.
- Open your scanned Santali image or PDF.
- Select Santali (sat) as the recognition language.
- Choose the appropriate page or region and run OCR.
- Review the extracted text, correct mistakes, and save it to a text file or another supported format.
If Santali does not appear in the language list, check the filename, the location of the model, and compatibility with your installed Tesseract version.
8. Method Three: Use Santali OCR on Linux
Linux is particularly useful for people working on language technology, research, and digital archives. The following example uses Debian, Ubuntu, or Linux Mint.
Install gImageReader and Tesseract
Open a terminal and run:
sudo apt update
sudo apt install tesseract-ocr gimagereaderThese commands install Tesseract and gImageReader from your distribution's available repositories. The installed version and package availability may vary.
Install the Santali language data
Download the compatible sat.traineddata model from the contribution linked above. Place it in the active Tesseract data directory. On many Debian-based systems, this directory is:
/usr/share/tesseract-ocr/5/tessdata/The path can vary by distribution and installation method. Check it using:
tesseract --list-langsIf sat appears in the list, Tesseract can find a language model with that code. This does not by itself prove that the model is the intended version.
Run OCR from the terminal
Place an image named santali.png in your current directory and run:
tesseract santali.png output -l satHere, santali.png is the input image, output is the output filename without its extension, and -l sat selects the Santali model. The extracted text is normally saved as output.txt.
For beginners who prefer a graphical interface, launch gImageReader and follow the same steps described for Windows.
9. What About Android ?
On Android, start with the online i2OCR service in your browser. You can also test Google Lens or the text-recognition feature of your preferred scanning application.
- Photograph the page in good lighting, keeping the camera parallel to the paper.
- Crop the image so that it contains the text you want to recognize.
- Open the i2OCR Santali image page in your mobile browser.
- Upload the image and follow the OCR instructions.
- Copy the result into a Unicode-compatible notes or document application.
- Compare the output with the photograph and correct character errors.
Google Lens is useful for general text extraction, but its results for Ol Chiki may vary. It should not be assumed to use the specialist Santali LSTM model. A dedicated Santali OCR service may be a better first choice for printed Ol Chiki pages.
For sensitive documents, consider privacy before uploading images to any online OCR service.
TextFairy
TextFairy is a powerful, free Android OCR app that converts images of text into editable formats.
Step-by-Step Guidelines:
Install & Add Language: Download TextFairy from the Play Store.
Open the menu, select Add Language, and download Santali. Capture or Upload: Tap the camera icon to snap a photo of the printed text, or select an existing image from your gallery.
Crop & Align: Crop the image to highlight the target text area, then proceed.
Run OCR: Choose the column count, select Santali as your recognition language, and tap Start.
Copy or Export: Review, edit, copy the extracted text to your clipboard, or save it as a PDF.
10. How to Improve Santali OCR Accuracy
- Scan printed pages at 300–600 DPI when possible.
- Keep the page straight and avoid shadows, glare, and motion blur.
- Use grayscale or a clear, high-contrast image.
- Crop unwanted borders while preserving all character marks.
- Test more than one page, especially if the book uses unusual fonts.
- Check missing characters, punctuation, spaces, and line breaks manually.
- Keep the original scan beside the extracted text for proofreading.
OCR output should always be treated as a draft. A good scan improves recognition, but it cannot compensate for every limitation in the model.
11. Can You Build Your Own Santali OCR Model?
Yes. Tesseract supports custom training, although building a reliable model requires technical knowledge and carefully prepared data.
A basic training workflow looks like this:
Collect Ol Chiki page images
↓
Prepare matching ground-truth text
↓
Check text and image alignment
↓
Train or fine-tune the model
↓
Test on unseen pages
↓
Correct errors and improve the datasetStart by collecting appropriately licensed printed material and creating accurate Unicode transcriptions. Separate training pages from test pages so that evaluation measures performance on material the model has not already seen.
Researchers can explore the training resources linked from Subhashish's Santali model contribution. Existing community datasets can provide a starting point, but every dataset's licence and reuse conditions should be checked.
12. Other Santali OCR Projects and Tools
- SantaliOCR on GitHub: an open-source project worth exploring for its code, documentation, and approach to Santali text recognition. Check the repository for its current scripts, requirements, and supported input formats.
- Tesseract OCR: a free, open-source OCR engine that can use language-specific trained data and supports custom training.
- i2OCR: a browser-based option for quickly extracting Santali text.
Commercial products such as ABBYY FineReader and PDNob PDF Editor may help with general document scanning and PDF workflows, but do not assume they recognize Ol Chiki accurately. Test them using representative Santali pages before choosing one for a large project.
13. Frequently Asked Questions
1. Is Santali OCR free?
Yes. Open-source tools such as Tesseract and gImageReader are free to use. Online services may impose usage limits or offer paid features.
2. What is the best tool for a beginner?
Try i2OCR first because it runs in a browser. Use gImageReader when you want more control over a local Tesseract installation.
3. Does Google Lens support Ol Chiki?
Recognition can vary. Test it with your document rather than assuming that it supports Ol Chiki as reliably as more widely supported scripts.
4. What does the code sat mean?
sat is the ISO 639-3 language code for Santali. It is commonly used to identify the Santali language model in Tesseract commands.
5. Is the new LSTM model completely accurate?
No. The initial reported results show that the model is experimental and needs improvement. Always proofread the output.
6. What is an LSTM model?
LSTM is a type of neural network that can process sequences. Tesseract's LSTM-based recognition approach is designed to recognize text sequences rather than relying only on older character-classification methods.
7. Can OCR read handwritten Santali?
It may recognize some handwriting, but models trained on printed pages usually perform less reliably on handwritten text.
8. Can I convert a scanned PDF into Word?
Yes, if your chosen OCR tool supports the required PDF workflow and export format. You may need to process individual pages and reconstruct the formatting.
9. Why does OCR produce boxes or incorrect characters?
The problem may be caused by poor image quality, an incompatible model, missing fonts in your editor, or recognition errors. Check both the model and the original scan.
10. How can I contribute to Santali OCR?
Test the model on different books, report reproducible errors, prepare accurate ground-truth text, contribute appropriately licensed scans, and share improvements with the project community.
14. Conclusion: Help Bring Santali Books into the Digital World
Santali OCR is more than a convenience: it is an important technology for preserving literature, improving access to educational materials, and developing digital resources for future generations. Although recognition accuracy still needs work, online OCR services and open-source tools give beginners a practical way to start.
Try i2OCR for a quick test, use gImageReader with the appropriate Tesseract language data for desktop work, and explore Subhashish's early Santali LSTM model. If you are a Santali speaker, teacher, student, researcher, or developer, your corrections and carefully prepared datasets can make a real difference.
Our next step should be to build better training data, test OCR on real Santali books, and share our findings openly. Every accurately digitized page helps make Santali knowledge easier to preserve, search, study, and share.
Reviewed by Prasanta Hembram
on
Wednesday, October 28, 2026
Rating:
Posts
No comments:
Please share your comments or feedback here.