When we opened the SciDraw AI API, we expected people to plug it into lab notebooks and slide tools. One of the first serious integrations went somewhere bigger: a whole book.
SciVolume is an AI book writer for academic books: monographs, textbooks and long-form scholarly guides. It plans an outline, writes chapter by chapter from sources the author approves, checks every citation and exports a Word, PDF or LaTeX manuscript. Every scientific figure in those books comes from our API.
Why a book is a hard figure problem
A journal article needs a handful of figures. A book needs dozens, written over weeks, across chapters on different subtopics, and they still have to look like they belong to the same book. Building and tuning an image pipeline for that is a project of its own. SciVolume chose not to build one and called ours instead.
How SciVolume uses our API
- The chapter writes the brief. While drafting, SciVolume turns the surrounding text into a short figure brief: what the figure must show, the labels it needs, the language of the book.
- One style per figure type. Each request asks for a schematic, a flowchart, a mechanism, a conceptual framework or a timeline, so figures in chapter 2 and chapter 9 share one visual language.
- Candidates, not a single shot. SciVolume calls
POST /v1/images/generationsand asks for several images at once. The author picks the best one instead of regenerating blindly. - Straight into the manuscript. The chosen figure is inserted with a numbered caption and shows up in the list of figures and every export.
Everything we have tuned for research figures, from the scientific-figure prompting to model routing and label accuracy, comes with each call. When we improve the figure engine, SciVolume's books get better figures without them changing a line of code.
You can see the result in their sample book, readable online with all its figures.
Build your own
The same API is open to any product that needs scientific figures:
- Generate figures from a prompt, at 2K or 4K, up to four per request.
- Edit an existing figure with a plain-language instruction.
- Convert between image formats.
- Track jobs and your credit balance.
Writing tools, electronic lab notebooks, course platforms, publishing workflows: if your users write about science, they need figures. Start on the developer page: create a key, run the sample request and check the credit cost of each operation.
Building something with the SciDraw AI API? Tell us about it; we would like to feature more integrations like this one.
