Starry Night program to create image gcode

thingiverse

This is a fascinating experiment in creative problem-solving. I have a friend Tim McGough who suggested that I attempt to recreate the Starry Night Van Gogh image by outputting colored dots using a swiping motion from left to right. This approach differs significantly from my previous method of producing STL files for each color, requiring custom gcode for the printer instead. I began by utilizing only black and white colors and creating a program that would output gcode to produce 1mm x 0.4mm lines of filament for each pixel, repeating this process for two or three rows at a time. Unfortunately, the result was an ugly, smeared mess with no discernible features other than the color transition. The diamond head's limited ability to switch colors quickly became apparent, making it necessary to allow more time between color changes. To address this issue, I modified the program to move 0.4mm up, then 0.4mm right, followed by a 0.4mm downward movement and another 0.4mm rightward motion for each pixel. This resulted in a 1.6mm dot per pixel, allowing more time for filament to emerge and improving the color accuracy. With several pixels of similar color on a single row, the image began to take shape, with shades of blobs corresponding to the Starry Night image visible. However, the process still produced a smeary mess, but at least it was now possible to discern the underlying image. I had some issues with x and y axis inversions in software, which I resolved by swapping them. Subsequently, I experimented with using 3x3 or 9 dots per pixel. The program would move up 0.8mm, right 0.8mm, down 0.4mm, left 0.4mm, and then down 0.4mm before finally moving right 0.4mm again for each pixel. This resulted in a 3.2mm per pixel size, which was larger than my print bed could accommodate, so I cropped the image using software. To my delight, this approach produced much more vibrant colors. However, my printer suddenly gave up on me, and I had to replace the control ramps board due to a faulty heated bed wire, which caused a current surge that blew out a mosfet, producing smelly blue smoke and melting some components. Despite this setback, I'm pleased with the proof-of-concept experiment's success. In the future, I would likely opt for a lower-resolution image to reduce the size and then use the 3x3 or 9-pixel method to achieve better color accuracy. To make the program more user-friendly, I'll need to incorporate custom printer settings, which will take some time. The program takes an input image, downsizes it to 160x104 with 12 black-and-white colors, and outputs a gcode file. The config input file isn't yet implemented, nor is the output color table fully utilized. To run the program, you'll need to install the GD module using "cpan install gd." Here's how to use the program: image2image.pl input_image_file config_file output_gcode output_color_table For now, this experiment serves as a showcase and likely the most expensive way to print an image, but feel free to play with the program at your own risk!