- Locate the executable: Find the MDpng2MDwebp.exe binary in the dist folder.
- Access the executable: Click on the dist folder to view its contents, including the .exe file.
- Download the executable: If you click on the file, you'll be taken to a screen where you can download the "raw" file by clicking the download arrow on the right-hand side.
Unsigned executable: Since the .exe file is not signed, Windows or Windows Defender may display a warning when you first run the program. You'll need to navigate through the warning prompts to find the option to run the program anyway.
Initial launch delay: Please be patient, as the program's initial launch may be slow.
Why do you want this program? If you have a Docling .md file or if you have any embedded b64 .png in an .md file, this will convert the image to webp, saving a tremendous amount of storage. You may be able to compress a PDF to be smaller than the original file, while keeping all the data you need for your use or for feeding an LLM.
In my testing by re-encoding the PNG stream with a variable sized webp stream at 20%, you can save up to 80% space. However, validate if graphics are good enough after processing.
You can use this for for Obsidian and should work on other .md files. Because I'll most likely never touch this again, the executable is attached, and will not be put on a release. If you download the binary, it should be completely self contained, and has no ability to change the config or the parameters.
If you run the program, it will present you with a dialog box, which is small and may not be focused. A picture of this box is below. You will need to hit "Open Markdown File" to activate a GUI file selector. You use the GUI file selector to find a file that has .png embedded images in it.
The program is slow to load at first because it is compiled python, but once loaded, is process and converts your current .md file page in less than a second.
I currently haven't purchased a cert, so you will see this when you run the program. If this is uncomfortable for you, then I would suggest running the python script, which will involve making sure you have all the dependancies loaded.
The program create a smaller version of the original file in the same directory and adds a _Webp to the end of the file name for identification.
It does not delete the original file, so if you not longer need it, you will need to delete the original.
My current process flow is to create markdown files with docling, which creates embedded png files After this file is created, you can run MDpng2MDwebp on the newly created file and it will convert the images
Savings is highly dependent on the original file and content, but here are three examples:
| Original PDF | Docling .md | MDpng2MDwebp | Savings from PDF (%) | |
|---|---|---|---|---|
| Owners Manual | 1,513 KB | 207 KB | 29 KB | 98.08% |
| Graphic Report #1 | 1,172 KB | 1,548 KB | 353 KB | 69.87% |
| Graphic Report #2 | 1,619 KB | 2,456 KB | 414 KB | 74.42% |