I explored raster drawing, vector drawing and 3D modelling. Raster images store pixels, which makes them useful for photographs and painted scenes. Vector files store paths and shapes, which are useful for scalable graphics and cutting outlines. A 3D model describes a volume that can later be exported for fabrication.
I made a sunset scene on a 2357 × 1308 pixel canvas. I used the line and fill tools to lay out yellow, orange and red bands, with darker red in the remaining area. To blend the horizontal colours, I reduced the canvas width to one pixel and enlarged it again. This produced a background on which I could draw the foreground.
I used black brush strokes for the mountains and tree, then a crayon brush for the leaves. I added a yellow sun and small bird shapes. The exercise helped me distinguish editing the image pixels from resizing the canvas, and showed how brush size changes the detail of a drawing.
In Pixlr, I imported an image, cropped it to a height of 1200 pixels and adjusted the colour balance. I added text, changed its colour and placed it at the top before exporting the edited image. The sequence below shows the crop, adjustment and text stages.
I also tried the AI image generator with the prompt “People talking in a café”. I selected and downloaded one of its outputs. This was an image-generation exercise; that output is not a photograph of work I performed.
I set up a 300 × 600 mm document and imported an animal reference image. I traced the outlines with the polyline tool, combined the paths, added a fill and placed a text label. I used alignment tools to centre the text and artwork, then exported a PDF.
Unlike the raster exercise, I could adjust individual nodes and paths without repainting the whole image. For fabrication, a closed, clean outline matters more than the appearance of the imported reference. The reference image is used here as a tracing exercise; its original source still needs attribution.
I created an A4 landscape document, 297 × 210 mm, and added “FAB ACADEMY 2026” using the text tool. I changed the text colour, drew a rectangle and aligned the elements. I saved the result as SVG so that the objects remain editable.
Both Inkscape and CorelDRAW let me arrange vector objects, but keeping the native or SVG file is more useful for later edits than keeping only a screenshot or flattened image.
I built a cup from simple solids. A cylinder formed the body, and a torus rotated by 90° formed the handle. I placed a second, smaller cylinder inside the body, set it as a hole and aligned the objects before grouping them. Subtracting that inner volume made the cup hollow.
I checked the handle position, changed the display colour and exported an STL. The STL stores the surface mesh for fabrication; it does not preserve the editable sequence of Tinkercad shapes.
For photographs, my workflow uses Squoosh with a width around 1600 pixels and MozJPEG quality near 80, checking the preview for readable labels and details. For video, I used HandBrake with the Fast 1080p30 preset, H.264 and an RF quality around 22. These are starting settings; the subject and text size determine how much compression is acceptable.
I have not recorded a before-and-after size comparison for these exports, so I do not assign a compression ratio to them. The aim is to retain useful process information while keeping pages practical to load.
The later enclosure work applies the same modelling ideas to a functional object: the display needs a located opening, and the solar panel needs a supported fit. The image below is from that project work. The assembly sequence is described in System Integration.
I learned that screenshots explain the steps, but editable files are necessary to change dimensions and reproduce a design. For the final enclosure, the original editable model still needs to be archived.
I am changing the final project to a 1.3-inch OLED because the original display is not available. The editable model for the earlier enclosure was missing from my archive, so I prepared a new Fusion design with Codex assistance. The script builds a native feature timeline rather than importing a finished mesh.
I started with a dimensioned 82 × 128 mm rectangle, extruded the base to 30 mm, rounded the vertical corners to an 8 mm radius and shelled it to leave 2.4 mm walls and floor. The rectangle sketches have horizontal and vertical constraints and dimensions linked to named parameters. This lets me change a dimension and recompute the model.

I added corner screw bosses and blind pilot holes, then supports for a removable tray. A separate 3 mm faceplate has a 31 × 17 mm OLED viewing window, display mounting holes and a 12.2 mm button opening. These are starting dimensions. I will replace the OLED hole pitch and button dimensions with the actual measurements before fabrication.

The native model includes the base, faceplate, tray, OLED backing and a screw-fit coupon. I exported the printable parts as STL and the assembly as STEP, keeping the .f3d as the editable design source. This new model does not replace the missing source for the earlier SolidWorks enclosure.
The CorelDRAW and SVG files retain editable vector objects. The cup STL is the exported 3D mesh.