First published 2026-09-15

tl;dr: no libraries that I tried convert images to text well, much less to Latex. I didn't try every library listed below.

Tesseract

Tesseract cannot natively produce LaTeX code from math equations or text.

citations:

• https://github.com/tesseract-ocr/tesseract/issues/4092
• https://github.com/tesseract-ocr/tesseract/issues/3028

How does Tesseract do with hand-written text?

sudo apt install tesseract-ocr
sudo apt install python3-pip
sudo apt install python3-pil
sudo apt install python3.14-venv
python3 -m venv .venv
pip3 install pytesseract

The script

from PIL import Image
import pytesseract
text = pytesseract.image_to_string(Image.open('20260628_090059.jpg'))
print(text)

produces

$ python3 convert.py  20260628_090059.jpg 
coatravarient vector
a= traa storm
Com poasn +5
ke dx (by A' ;)
covarient vec tof Componsr tS
ror storm )
A stor like et (by B, se
de; ar of de! describe the same
physceal thing (di tlre ntal elerrnt of
— leagth or oli tener thal element oF angle )

LaTeX-OCR

https://github.com/lukas-blecher/LaTeX-OCR aka https://lukas-blecher.github.io/LaTeX-OCR/ Developer is the lead author on the paper "Nougat: Neural Optical Understanding for Academic Documents" (https://arxiv.org/abs/2308.13418)

Nougat: Neural Optical Understanding for Academic Documents

https://arxiv.org/abs/2308.13418

From Facebook Research:
https://github.com/facebookresearch/nougat aka https://facebookresearch.github.io/nougat/

im2latex.py

pix2tex

https://pix2tex.readthedocs.io/en/latest/

deprecated: https://github.com/blaisewang/img2latex-mathpix

Local LLM

Serve the LLM using

python3 -m llama_cpp.server --model ./models/gemma-4-E4B-it-qat-UD-Q4_K_XL.gguf --n_gpu_layers -1

curl -X POST "http://localhost:8000/v1/chat/completions" \
-H "Content-Type: application/json" \
-d '{
  "model": "gemma-4-E4B-it-qat-UD-Q4_K_XL.gguf",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Perform OCR on this image. Extract and output all text visible in it precisely."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "data:image/jpeg;base64,'$(base64  image.jpg)'"
          }
        }
      ]
    }
  ]
}'

The output does feature Latex, but nothing related to what's in the image.

Appendix: a script that converts HEIC to JPG

convert_heic_to_jpg.py contains

import argparse
from pathlib import Path
from PIL import Image
import pillow_heif

# Register HEIF decoder with Pillow
pillow_heif.register_heif_opener()

# Set up command-line argument parsing
parser = argparse.ArgumentParser(description="Convert a HEIC image to JPEG.")
parser.add_argument("input_image", type=str, help="Path to the input HEIC image")
args = parser.parse_args()

# Handle input and output paths using pathlib
input_path = Path(args.input_image)
output_path = input_path.with_suffix(".jpg")

# Open and convert the image
try:
    image = Image.open(input_path)
    image.save(output_path, "JPEG")
    print(f"Successfully converted '{input_path}' to '{output_path}'")
except Exception as e:
    print(f"Error converting image: {e}")