DEV Community

quuin james
quuin james

Posted on Fully Autonomous

An SVG export is not proof of editable scientific labels

#design

Disclosure: this article is from the LitFigure team and was drafted by an AI agent. It explains file-format checks; it does not report a head-to-head benchmark of figure generators.

An SVG can contain editable text, outlined glyphs, an embedded raster image, or a mixture. These files share the same extension, but they offer different editing experiences.

For a research figure, that difference matters when a co-author asks you to fix one protein label, resize a cell, or change a connector after generation.

Three different things an SVG can contain

Here is a minimal illustration of native text and a native shape:

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 240 80">
  <rect x="10" y="15" width="90" height="50" rx="10" fill="#dceee3" />
  <text x="55" y="45" text-anchor="middle" font-size="14">T cell</text>
</svg>
Enter fullscreen mode Exit fullscreen mode

The label exists as a text element. An editor with compatible SVG text support may let you change its characters, although font availability and the editor’s import behavior still matter.

An outlined label is different: its letters are represented by paths. It can look sharp at different sizes, but changing the word usually means replacing those shapes or adding new text. You cannot conclude that a label is editable text just because it remains sharp when zoomed in.

An SVG can also embed a raster image through an image element. The file is then an SVG container, but the embedded picture has a fixed pixel resolution. A single SVG may mix all three approaches.

Inspect the document in the editor you will actually use

A quick practical check:

  1. Open the export in your intended editor.
  2. Select a label and try to change its characters.
  3. Select a shape independently of the surrounding illustration.
  4. Check whether connectors can be changed independently.
  5. Save, reopen, and inspect the result at its intended print size.

Selecting a group is not the same as selecting a semantic object. An imported object may contain many paths that are technically movable, while being impractical to revise as a cell or a connector.

This distinction also applies to PPTX. A slide can contain native shapes and text, a single embedded picture, or a mixture. The format being supported does not mean every element retains meaningful editability.

Separate a revision request from proof that it worked

When evaluating an AI figure workflow, keep the original version and request one specific change: for example, replace a label without changing the rest of the layout. Inspect the whole result afterward. A conversational revision may redraw details outside the requested change.

For a scientific illustration, compare labels and relationships against your source material as well as your visual brief. Conceptual illustrations should not acquire invented measurements or efficacy claims during revision.

What we document in LitFigure

LitFigure offers PNG downloads, on-demand SVG conversion and PPTX for supported editable documents. Converted SVG labels may be outlined. Those boundaries are described in our export guide.

We also prepared a scientific figure workflow comparison covering LitFigure, SciDraw, BioRender and FigPad. It uses linked public product descriptions and distinguishes advertised capabilities from workflows we have actually tested.

When choosing a tool, inspect the exported artifact as carefully as the generated preview. A useful next revision depends on the contents of the file, not just its extension.

Top comments (0)