Released v0.4.2 as a patch release for stabilizing OCR text extraction, with an addition to multiple fonts usage (you can checkout the 2 previous devlogs for more)

Changelog

• Release v0.4.2
• misc: avoid using namespace spdlog;
• args: add –override + font: add default if empty string
• config: improve fonts acquiring
• misc: some coding stuff

Included support for multiple fonts usage, which can be necessary if the default one doesn’t support the extracted language

Changelog

• Add support for multiple fonts (#6)
• misc: just check if font path exists

Fixed detection of some special characters, and also improved a bit the UX (the font in the video isn’t be the default one).

Changelog

• ocr: offload more guesswork to tesseract (#5)
• tool: improve error view + add details tree on ocr
• misc: forgot to remove from translation purge
• ocr: return grayscale from preprocess_pix instead of 1bpp

oshot v0.4.0 has been released!

Will soon make ship, I just wanna test if things works just fine as a first impression

fixed bogus transparent background, and also fixed error sets when capturing from systray

Changelog

• tool: ocr: fix refresh of models for retriving potential errors
• idk
• render: fix bogus transparent image background
• annotation: add CTRL+Z undo annotation shortcut

made text annotation input color the same as when placing it. This can be useful if typing in certain backgrounds colors, for example entirely white ones and you can’t see what you typed

This devlog is used for reporting every bug fixes I have fixed since v0.4.0-rc1 release.

Changelog

• appimage: try to fix linking error from some missing libraries in bundle
• windows: fix auto-quit after capturing from systray
• log: use our own fmt bundle
• log: complete implementation with oshot.log + fix compile
• workflow: continue other OSes build even 1 fails
• clipboard+socket: remove htonl() and ntohl() + fix threading and guard clipboard sender on connection state
• log: implement spdlog (WIP)
• tool: fix stale when opening a new image while having an ann. active
• Release v0.4.0-rc1
• glfw: put the RAII guard in util.hpp
• glfw: glfwTerminate() everywhere

added option to not render annotations in Image scans (OCR, QR/Barcodes text extractions)
modifiable in the config.toml file too

Changelog

• config: add option for optionally rendering anns. in image scans
• metal: remove some OpenGL only functions
• capture: fix multi-capture on macos and windows
• capture: make KDE a separate session
• menu: about: add copy text button

Improved the Help->About window, with logo, centered text and a “More Details” for giving informations when reporting bugs/issues

I wanted to put in this devlog the changes I have done which are mostly fixes on mutli-monitor setup captures, new PRs being created and merged (thanks for the interest!) and other fixes for the new upcoming v0.4.0 release.

Stay tuned!
the image below was taken when there was an issue with multi-monitor setup capture, that’s now been fixed!

Changelog

• capture: try to fix on KDE plasma
• capture: remove portal capture + fix (hopefully) multi-monitor capture
• build: fix
• socket: remove lock at exit
• tool: capture only the focused monitor + other things
• glfw: fix ignored hints
• systray + socket: use AF_UNIX (#3)
• wayland: fix hanging with every text/image copy (#2)
• windows: trying to fix exit lag issue on some machines
• annotation: settings: make the loupe crosshair centered
• chore: rename main_tool_glfw.cpp -> main_tool_opengl3.cpp
• chore: some variables re-names
• Update README.md

Translation has been purged from oshot.
It was a PITA trying to make it work and stable without API keys and subscriptions. The only solution would be to use libretranslate, but I doubt anyone would host a kinda heavy server for just making a screenshot tool translation work.

But as William Afton says: I always come back

Changelog

• text tools: purge translation and anything related

Scale the input text based on its font size

Though the default font (ProggyClean) won’t work since it’s not meant to scale all

MacOS support has arrived

Oh my God. What a journey.
Having to refactor from GLFW to Metal just for macOS was a great PIA. I’m glad the results has shown off though.

It is on a separate branch and will be merged to main after some issues have been fixed. Soon will post the video demo too.

Shootout to @Hridya for testing my program on his mac and allowing me to post a devlog using his screenshot as a preview.

btw I think this took MORE than just 5h

Now you can open/close the text tools window

changeble by config too

Changelog

• config: make text tool window closable/openable

Fixed dragging selection

Fixed text and other annotation rendering with colors

In the image from the previous devlog, the compression algorithm made the text rendering super dirty

Implement text annotation

Changelog

• tool: make shortcuts work outside of the text tool window
• annotations: text: fix not rendering in image
• chore: implement error_v() and get_v() in Result<>
• annotations: text: implement custom font settings + some codebase changes
• annotations: add text

Upgraded the color picker to next level

Changelog

• annotations: refactor color picker to next level
• args: integrate back –tray for more control
• tray: fix socket lock
• chore: rename std::filesystem -> fs
• windows: fix compilation and crash
• main: switch –tray to –gui

Image 1: Implement color picker

Image 2 (now) and 3 (old): Fix PSM autodetect + improve preprocess for a better OCR result

Changelog

• annotations: implement color picker
• annotations: forgot to implement filled rectangle and circle in GetFinalImage()
• ocr: improve preprocessing for dark-bg and RGBA images
• fix incredibly stupid bug

Shipped first bug fix release!

Changelog

• refactor render logic
• Release v0.1.0
• introduce settings
• silence some warnings
• Update README.md

Introduce settings

fixed font path search

Changelog

• fix font path search
• remove socket logic on windows since it’s useless
• re-integrate back the socket sender

finalized wordle game (sorry for non-centered recording)

Changelog

• make every scene/game render() blocking by default
• finalize wordle
• fix colors on windows and wordle
• release beta2

implementing wordle, still needs to add scores and errors and such

cliboy has been ultra debloated

we did it

I see….. rewritting from scratch by hand without AI fixed the issue….. Imma removethe AI usage tag soon maybe
(yes I know the blue thing, fixing it)

wanna be honest, the hours I have logged in this devlog has been just a waste of time for trying to rewritte the TerminalDisplay API (the one that is used for rendering fonts, texts and colors) from notcurses to termbox2.h, which it was giving MASSIVE issues when trying to set a terminal font. No amount of usage from ChatGPT, v0.app and deepseek has helped me at all.

this meme summarizes what I did in this hour.
I’m sorry windows users but I’ve gone through hell with cmd.exe

Refactored codebase to be more modular and easier to add games

time to get back to work on this

Changelog

• add filled rectangle and circle
• improve annotations render + some perfomances

made thickness variable per tool, also added some new options such as the alpha ones

Introduced annotations

now my program can be an alternative to bloated flameshot

Changelog

• refactor icons
• introduce annotations
• make window actually fullscreen
• fix displaying multiple errors

Added delay (ms) for capturing screen

program so fast that they told me to slow it down lol

dev stuff: added better result return in functions. instead of unknown failed errors, let’s actually be a bit more precise (a bit buggy ik, fixing it when I want. edit: fixed)
also added the —debug flag in releases build too for reporting issues

oshot v0.3.0 finally released!

Added manual folder selection, with drag-n-drop too

also fixed various crashes related to keybinds and others

Refactored the system tray in native format

Thus python is no longer required

added result details

such as confidence score, and symbology of detected barcode
shootout to https://qr.bgtulk.dev/

Fixed the window hovering incosistency

Add Open Image option

I wanna also thank everyone who voted for my project, the next ship will happen soon.

decided to make a bunch of code cleaning and reformatting, along side adding the portal screenshot capture as fallback on linux. Overall, these changes have been just some fixes such as the refactor of DXGI capture on windows, and code style reformatting. Once the payouts are developed, i’ll add some features and release with a new ship

Added QR/Bar code support

most of the time was spent on some bugfixes and releases, this implementation took like 30mins if not less thanks to zbar

Add option to override PSM auto detection

Also made lots of improvements and fixes to the perfomance and security of the app

Improved OCR quality based on the image size

Resizing/Moving selection added

background song is elite ball knowledge

Now you can open oshot with other images

Even from stdin

added shortcuts + copy image
also reworked the network listener

added mini launcher and copy text

Should’ve been here since before

fixed multi-line translation + add option to edit OCR output

New Menu bar

Improved error handling + continuos checks

Thus improved a bit the perfomances too

Better Translation support + Custom Font based on language

ditched the bash script for better perfomance and support on both windows and linux

Custom fonts support added

Windows Support on the way

Misc: configuration for default settings + cli –arguments

Added Translations

(todo: different languages font support to be added)

added custom settings (implementing saving these soon)

fixed window padding

Local OCR with Tesseract

has flameshot finally come to wayland?

managed to aquire x11 fullscreen screenshot, testing to wayland later. The PPM saving is just a test.

still learning glfw+opengl3+imgui