Skip to main content
MIP can draw publication-quality scientific figures on demand. Ask for a pathway, a labeled diagram, a multi-panel infographic, or a clean illustration, and a rendered figure appears inline in the chat. No separate tool, no prompt-engineering boilerplate, no copying into PowerPoint afterwards. A scientific illustration generated inside MIP

When to use it

Figure generation is for things you would normally open BioRender, Illustrator, or a slide deck to produce:
  • A labeled diagram of a pathway, receptor, complex, or cell
  • A multi-panel infographic summarizing a workflow or experiment
  • A schematic of a mechanism of action
  • An illustration of a protein, organelle, or signaling event
  • A visual summary of results you just pulled from databases in the same chat
For tables of numbers, charts of data, or step-by-step text explanations, use the regular chat response, a table, or a chart instead. Figures are for visual concepts, not structured data.

Asking for a figure

Simply describe what you want:
  • “Draw a diagram of the MAPK/ERK signaling pathway with the key kinases labeled.”
  • “Create an infographic summarizing the three-step workflow I just ran.”
  • “Illustrate a nucleus with the major compartments annotated.”
  • “Make a schematic of antibody-antigen binding at the CDR loops.”
  • “Visualize the VEGF signaling cascade as a pathway figure.”
The more specific you are about the subject, components, spatial layout, and labels, the better the output. A one-line prompt like “draw a cell” gives a generic result; a paragraph naming the exact components, relationships, and labels gives a figure you can drop straight into a paper or slide.
Put any label text you want preserved verbatim inside double quotes. MIP passes quoted strings through to the renderer unchanged so they show up correctly on the figure. Example: labels “IL-6R”, “JAK1”, “STAT3”.

Intent, size, and quality

MIP exposes a few controls you can set explicitly if you want a particular look. Most of the time the default (auto) is correct and you can ignore these.
ControlValuesWhen to set it
intentdiagram, infographic, schematic, pathway, illustration, autoSet when you want a specific figure style. pathway emphasizes directed arrows; schematic favors clean line art; infographic lays out multiple panels.
sizesquare, portrait, landscape, autoSet when the figure needs to fit a specific layout (a poster panel, a slide header, a paper column).
qualitystandard, high, autohigh for dense text, many labels, or final-quality output. standard for quick drafts.
backgroundwhite, transparent, autoUse transparent when the figure will sit on a colored background (slide, poster). Default is white.
You can also write a natural-language request and MIP will pick sensible defaults. Mentioning “multi-panel” nudges it toward landscape and high quality; mentioning “quick sketch” nudges it toward standard.

Building figures from earlier chat results

The most useful pattern is asking for a figure that synthesizes something you just did. MIP has the full context of the conversation, so it can lift the right elements into the diagram. Examples:
  • After a deep-research query on PCSK9: “Draw a diagram of the PCSK9-LDLR mechanism based on what you just found.”
  • After a protein visualization for BRCA1: “Make a schematic showing the BRCT and RING domains we just discussed, with key residues labeled.”
  • After an enzyme design result: “Illustrate the designed active site with the substrate in place. Label the candidate catalytic residues.”
MIP pulls the relevant facts, molecules, residue numbers, and labels from the preceding turns and produces a figure grounded in that material rather than a generic illustration.

What you get back

The generated figure appears inline in the chat as a rendered image. You can:
  • Right-click → Save image as to download the PNG locally
  • Click to open in a full-size preview
  • Ask for a revision in the next turn (“change the arrows to dashed”, “move the substrate to the top-left”, “relabel ‘Gene A’ as ‘BRCA1’”) and MIP will regenerate with your edits applied
Every figure is also persisted, so coming back to the chat days later will still show the image.
Figure generation is a creative tool. For anything destined for a paper, a regulatory submission, or a clinical report, the figure should be reviewed, edited in a dedicated graphics tool if needed, and verified against the underlying data. Use it as a first draft that saves you the BioRender hour, not as a final deliverable.

Tips for great figures

  • Describe, don’t list. “A kidney nephron showing the glomerulus, proximal tubule, and loop of Henle, with blood flow arrows and filtration labels” beats a bare bullet list.
  • Anchor spatial layout. Say “top-left”, “flow from left to right”, or “two panels side-by-side” if the layout matters.
  • Quote exact labels. Wrap every term you want rendered verbatim in double quotes, especially gene names, residue identifiers, and abbreviations.
  • Start general, then refine. A first pass establishes the composition; follow-up turns polish labels, colors, or emphasis.
  • Lean on context. If you have run searches, lookups, or analyses in the same chat, ask the figure to draw from them rather than re-specifying the content.
  • Pick intent deliberately for dense diagrams. infographic is the right default for a multi-panel summary; pathway is the right default for signal flow with arrows; diagram is the right default for an anatomical or structural schematic.
Figures are a natural companion to Deep Research. Ask a question, let MIP synthesize the answer from 40+ databases, then ask for a figure of the result. The figure inherits every fact MIP just pulled.