Here is a pyproject.toml example

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[project]
name = "textshot"
version = "0.0.1"
authors = [
  { name="Ian ianzhao05", email="[email protected]" },
]
description = "Python tool for grabbing text via screenshot"
readme = "README.md"
requires-python = ">=3.7"
classifiers = [
    "Programming Language :: Python :: 3",
    "License :: OSI Approved :: MIT License",
    "Operating System :: OS Independent",
]
dynamic = ["dependencies"]

[project.urls]
"Homepage" = "https://github.com/ianzhao05/textshot"
"Bug Tracker" = "https://github.com/ianzhao05/textshot/issues/"

[project.scripts]
textshot = "textshot.textshot:main"

[tool.setuptools.dynamic]
dependencies = {file = ["requirements.txt"]}

[tool.setuptools.packages.find]
where = ["textshot"]

It may need some tweaking and modifying the project to relocate python files in textshot and changing the way import is done by using something like from .ocr import … for example.

It would be helpful to make a working package for linux distributions.

You then build and install the package with:

python3 -m build --wheel
python -m installer dist/*.whl

Does this repo support Chinese?

When I run textshot opposite this problem. I show in video. Sorry bad english if I mistake anywhere.

I tried to run textshot on my fedora Linux machine and I got a segmentation fault error:
Traceback (most recent call last): File "/home/maerqin/PycharmProjects/Screenshot_To_Text/venv/lib/python3.12/site-packages/textshot/textshot.py", line 11, in <module> from .logger import log_copied, log_ocr_failure ImportError: attempted relative import with no known parent package [1] 82030 segmentation fault (core dumped) python textshot.py -h

Please add a license file.

Sometimes the target is too small on the screen and I can't capture it accurately. Maybe it is a good idea to add a magnified image based on what is around the cursor when users are capturing the screen.

When I make a selection starting not from the upper left corner, but from any other corner, instead of the text in selection it returns some long random text, which seems to be the text from all screen.

To reproduce, make a selection of some text from, for example, lower right corner to upper left corner.

Everything works except the starting window when textshot is called, which is fully grey.
OS : Manjaro Linux KDE

Here is an example.
The positions captured by the program:

start:1227,695
end:1272,715

Their real positions:

start:3681,2085
end:3816,2145

Thus pyscreenshot grabs the wrong image.

My scale factor:

GDK_DPI_SCALE=0.333
GDK_SCALE=3
QT_AUTO_SCREEN_SCALE_FACTOR=0
QT_SCREEN_SCALE_FACTORS=eDP1=3;DP1=3;DP2=3;HDMI1=3;HDMI2=3;VIRTUAL1=3;

Use the textShot.ahk script that comes with it
Please guide

ERROR: An error occurred when trying to process the image: (1, "Tesseract Open Source OCR Engine v3.05.00dev with Leptonica read_params_file: Can't open txt Warning in pixReadMemPng: work-around: writing to a temp file libpng warning: Application built with libpng-1.4.3 but running with 1.5.14 Error in pixReadStreamPng: png_ptr not made Error in pixReadMemPng: pix not read Error in pixReadMem: png: no pix returned Error during processing.")

On recent Arch linux with i3wm window manager I often get ERROR: Unable to read text from image, did not copy.

I have :

• python-pyqt5 : 5.15.8
• python-pyqt5-sip: 12.11.1
• python-pillow: 9.4.0
• python-pytesseract: 0.3.10

This tool could only take a shot on its current desktop. However, MacBook has a multi-desktop feature, and you can't ask this tool to take a shot on the desktop where the terminal is opened. Hope the author can support multi-desktop screenshots.

Hi
I think this is a really cool idea to make OCR simple to do and allows for correcting OCR mistakes very easily.
I am on a Windows machine and I find that I need to OCR a large text image in parts because it doesn't handle
a lot of text well. Is there a recommended maximum amount of text that should be selected for conversion?
But even doing OCR in parts, some areas appear to be captured and the "spinning wheel" indicates that
a conversion is being done. But when pasting the text that is in the clipboard to notepad++, it is the text from a prior conversion.
If there is an error in the conversion process, I can't find where it is displayed. Can you please give me some pointers
on getting around these issues?
Thanks!

很遗憾

Anyone encountered this phenomena?

好项目呀，感觉很实用，不过我看现在应该是只有英文，之前chineseocr有17M中文识别模型模型，
还有最近百度飞桨新发的，https://github.com/PaddlePaddle/PaddleOCR 只有9M的模型，效果好像还更好一些，
不知道几位大佬，最近有没有计划把中文识别能力集成进去呀？

I have two screens (let's name them Main and Side). When I open type textshot in a terminal in Side, the Main monitor starts mirroring Side monitor's content.

So, to copy text from Main, I have to open the terminal in Main. This is not a good experience

The screen overlay does not pause the desktop while snipping. For example, videos and GIFs continue to play in the background. This is inconsistent with Windows's screenshot tools, Snipping Tool and the newer Snip & Sketch.

Raised by @rigred in #12

I use two monitors , this program can't work

Just a friendly heads up that I've added your package to the archlinux aur and it will keep itself updated based on the latest git commits to the github repo.
https://aur.archlinux.org/packages/textshot-git

So for arch users it's as easy as installing textshot-git with their favourite aur helper.
yay -S textshot-git

Also is there a way to make textshot pause the desktop (animations like the gif on this page)?
Currently it keeps on animating while in box select mode.

Thanks for the great tool!

Cropping is incorrect at times. It depends on screen resolution. I have a 4k display and it wasn't performing as expected

I just downloaded text shot on my Mac and installed all the dependencies but have been experiencing this weird behavior where as soon as I run it it will open a new screen to the right with no open apps and would only allow me to screenshot there. Did anyone else encounter this or have a fix?

2021-02-27 19:08:54.739 Python[952:13732] ApplePersistenceIgnoreState: Existing state will not be touched. New state will be written to /var/folders/ld/wjmpqdpj1pq2j4j_svh1j8740000gn/T/org.python.python.savedState

I have created a shortcut in my Ubuntu for textshot and whenever I use it (some times not all the time) the text is copied in French and Not being copied to clipboard at all(this is main issue), I knew it was in French because it was shown in notification :/

Once I unplugged from external monitors it worked well. The problem is raised from line 29 in textshot.py,
self.screen = QtWidgets.QApplication.screenAt(QtGui.QCursor.pos()).grabWindow(0)

My system is MacOS 10.15.3, with python 3.6.9. Thank you.

Hi,

On macOS 11.1, invoking python textshot.py throws a Qt GUI error:

QPixmap::fromImage: QPixmap cannot be created without a QGuiApplication
QPixmap: Must construct a QGuiApplication before a QPixmap

Any suggestions? This worked fine before updating to Big Sur. Thank you!

EDIT: This has been tried with a virtual environment.

screenshot "E:>cd github" output -->"INFO: Copied "AR" to the clipboard"

the command line test info:

E:\github>cd..

E:>cd github

E:\github>cd textshot

E:\github\textshot>python textshot.py
**INFO: Copied "E:\github>cd. .
AR

E:\github>cd textshot" to the clipboard**

E:\github\textshot>python textshot.py chi_sim
**INFO: Copied "ET
E:N>cd github

E:Ngithub>cd textshot" to the clipboard**

E:\github\textshot>

Appears to not work on more than 5 words at a time, presents with error

"TextShot"
"An error occurred when trying to process the image: Tesseract process timeout"

on Mac get

INFO: Unable to read text from image, did not copy

seems pyperclip do not work properly on MacOS.

Hello the issue is my brain, I can't make it work pls help.
Basically, I can make textshot work via cmd but am too dumb to understand the greatness of your coding skill and btw how autokhey works.
Pls help