I’ve occasionally used ImageMagick’s erode and dilate filters to make art look a little less digital. Turns out those filters are also available in Photoshop and Affinity Photo, just under different names: minimum blur and maximum blur. I had no idea. (I should mention that there might still be subtle differences between the algorithms. I didn’t do a deep dive. But from my limited testing they seem to do the same thing.)
Composite: convert go-with-him-twain-bg.png \( go-with-him-twain-eroded.png -normalize +level 0,55% \) -compose screen -composite go-with-him-twain-composite.png (I usually do this in Affinity Photo but I wanted to try using Imagemagick)
Upscale to 8000×10000px with Real-ESRGAN: ./realesrgan-ncnn-vulkan -i go-with-him-twain-composite.png -o go-with-him-twain-full.png -s 2
Add noise (8% monochromatic) in Affinity Photo and export final PNG (I could do this in Imagemagick, need to port it over)
I’m experimenting with a slightly new style here, masking the notes (after adding noise to the outlines with SVG filters and then eroding/dilating with Imagemagick) and painting inside the mask in Procreate. (And then texturing it in Affinity Photo as usual.)
Typeset the first phrase (or so) in MuseScore using the Bravura font, with the spacing trimmed to within an inch of its life
Play it out loud to make sure I entered it right (cough) and export an SVG
Drag the SVG into a frame in Figma and use the SkewDat plugin to skew it -4°, center it, then export a 4,000px-wide PNG
Use ImageMagick to do some erosion and dilation (to simulate age and ink spread): convert input.png -morphology erode disk:18 -morphology dilate disk:16 output.png
Texture the image in Affinity Photo and export the PNG
Upscale with Real-ESRGAN to 12,000px-wide
Downscale a little in Affinity Photo, add 8% monochrome noise, and export the final PNG
I’m still figuring out how I want to do these (full bleed or not, barlines, clefs and key signatures or not, etc.). Also thinking about possibly doing some abstract versions as well, to avoid all these music typesetting issues entirely.
I’ve recently begun scanning my journals using my iPhone and the Scanner Pro app, and it’s working out fairly well. My process:
Using the built-in iPhone camera app, I long press to lock focus and exposure (this saves time so it doesn’t have to autofocus each time), then photograph each page of the journal. It’s not as high quality as it would be if I used an actual scanner, but it’s much, much faster, and far more portable.
After I’m done photographing, I open Scanner Pro and select the images from the camera roll, then use the Black & White Document setting to process them into a PDF.
From Scanner Pro, I export the PDF to Dropbox.
The resulting PDF is nice and clean and easy to read, and the files aren’t too big (150 pages is usually between 80 and 200 megs — for me, very much worth the space to preserve important documents).
A concocted example:
That’s before (the image is straight from my iPhone camera, no postprocessing), and this is after Scanner Pro is done with it:
I should add that ordinarily, with actual journals there wouldn’t be as much empty border around the content.
One hitch I’ve run into is that Scanner Pro chokes on anything larger than around 150 pages (it crashes), so I do long journals in chunks.
For that reason and a few other small annoyances, I’ve been looking into replacing Scanner Pro with a desktop-based script that takes a list of photos and processes them into a nice black and white PDF. Imagemagick gets me part of the way there with this command:
Migrating to Day One has resurrected my efforts to scan and transcribe my older paper journals. As I’ve been doing this, I’ve run into the need for a couple small shell scripts to automate things.
On several of these journals I’m scanning the full two-page spread because the whole journal fits on the scanner platen, which means splitting the resulting image out into two (one for each page). Splitimage uses ImageMagick to do that nicely. There’s some overlap, but for a fully automated solution it’s not bad, and it saves me a lot of time cropping.
I prefer taking these split images and renaming them sequentially using something more meaningful (“journal-2009.005.jpg” rather than “IMG_0034.JPG”, for example). I used to do this with OS X’s Automator tool, and it works quite well, but I wanted a quick command-line tool to speed things up. Enter dub, a zsh script that simplifies the batch renaming process. Now I can just type:
dub journal-2009.X.jpg *.JPG
And then it’s just a matter of dumping them into Unbindery and transcribing them.