Getting Started With Science Things To Draw
Science Things To Draw is a browser-based application designed for creating precise scientific illustrations — anatomical sketches, molecular structures, geological cross-sections, and similar technical diagrams. It started as an internal tool at a small research lab and eventually got open-sourced a few years ago. The codebase lives on GitHub, and the project page gives you a direct download or can be run locally if you clone the repository. The core idea is straightforward. You start with a blank canvas, pull in reference images or coordinate data, and build up vector-based drawings using a set of specialized tools — bezier curves, grid snapping, scale bars, and annotation layers. What sets it apart from general drawing programs is the built-in support for scientific conventions: proper labeling hierarchies, scale calibration against known measurements, and export formats that play nicely with journal submission requirements.
My experience with the rendering pipeline
When I first started using this tool, I ran into a problem with vector export quality at high zoom levels. The default SVG renderer was dropping path details when the drawing contained more than about two hundred bezier curves. My cross-section diagram had around four hundred curves because I was working from a scanned lithology sheet, and the exported file came out jagged and incomplete in any program that opened it. The workaround was to set the rendering engine to "precision mode" in the settings before exporting, which doubles the internal node resolution and produces a much larger file but preserves every detail. It took the export time from about three seconds to roughly forty-five seconds, but the output was clean. Another thing that trips people up is the coordinate system. The default grid uses arbitrary pixel units, not real-world measurements. If you are doing something where accuracy matters — molecular bond angles, microtome slices, anything that needs to match published data — you need to calibrate the grid using a scale bar reference before you start drawing. I skipped that step on a protein structure project once and had to rebuild the whole thing after my collaborator pointed out the bond angles were off by about twelve degrees because the grid was uncalibrated. Took me twenty minutes to redo it properly.
What it does well and where it falls apart
The annotation system is genuinely useful. You can layer text labels with leader lines, attach measurement callouts, and nest them in groups that toggle visibility. This matters when you are preparing figures for a paper and need multiple versions — a simplified label-free version for the main text, a detailed version for supplementary material. Setting that up used to take me about an hour in other tools. With this, it is more like fifteen minutes once you get the layer organization down. The molecular structure templates are decent but not comprehensive. Standard organic molecules, amino acids, and common minerals are all there. If you are working with something niche — unusual coordination complexes, obscure crystalline polymorphs — you will be building those from scratch anyway. The curve tools handle it fine, but you lose the convenience of pre-set bond lengths and standard angles that other dedicated chemistry drawing programs offer. For most biology-focused work this is not a problem. For chemistry or materials science, you might be better off using something like ChemDraw or a dedicated crystallography visualization tool and importing the result as a reference image into this app. Performance degrades noticeably when you are working with imported raster scans at high resolution. I had a thirty-megapixel photomicrograph that I wanted to trace over, and the application started lagging badly around the two-hundred-megapixel equivalent mark. The workaround is to downsample the reference image to something reasonable — six hundred to eight hundred pixels wide is usually plenty for tracing purposes — before loading it into the canvas. The app does not do this automatically, and it will not warn you.
Get the Full Details
Practical workflow
Start by setting your canvas size and establishing a scale. Use the grid calibration tool and input the real-world measurement for your reference frame. Import any source images or data files you need. Build your drawing layer by layer, keeping related elements grouped. Add annotations last so they do not get tangled in the geometry work. Export to SVG for editing flexibility or to PDF for direct submission to most journals. The file sizes tend to be on the larger side for detailed work — expect ten to twenty megabytes for a complex illustration — so plan your storage accordingly. The download is available through the project's official GitHub releases page, and there is also a web version that runs in the browser without installation. If you are on an older machine or have limited local storage, the browser version avoids the performance issues I mentioned above for lighter projects. For heavy work with large scanned references, running it locally gives you more control over the rendering pipeline and tends to be faster overall.
