prompt

Free

Expert paper-figure illustrator prompt for matplotlib style mirroring

FreeFree tier

Type

Open Source

Visit Website GitHub (40 stars)

About prompt

Paper Figure Mirror Engineer is an advanced AI prompt from the awesome-prompts collection, designed to transform an AI into an expert paper-figure illustrator using matplotlib. The prompt instructs the AI to mirror the visual style of a reference figure (e.g., palette, spine treatment, gridline weight, marker shape) onto the user's own data, while respecting geometric reservation, palette fidelity, and typographic restraint. It includes a multi-stage process: reference preprocessing (cropping and cleaning), aesthetic convention library (L2) for low-resolution references, iterative refinement with configurable iterations (default 6), and a 'ship' verdict when quality passes. The tool accepts data in various parseable formats (CSV, TSV, markdown tables, pasted tables, dirty terminal text) and a reference image. The output is a matplotlib figure aimed at being indistinguishable from a hand-tuned figure by a senior author of a top-tier ML/CS paper. This prompt is open-source and available for free on GitHub.

Key Features

Mirrors visual style (palette, spine, gridline, marker, legend) from a reference image

Automatic cropping and cleaning of reference image (Stage 0)

Iterative refinement with configurable iteration count (default 6) until quality threshold is met

Aesthetic convention library (L2) for handling low-resolution or ambiguous references

Parses user data from CSV, TSV, markdown tables, pasted tables, and dirty terminal text

Produces publication-ready matplotlib figures for academic papers

Pros & Cons

Pros

Produces high-quality, professional figures that mimic top-conference styles
Accepts a wide variety of data input formats
Includes iterative refinement to ensure output quality
Open-source and free to use
Detailed prompt instructions reduce ambiguity for the AI

Cons

Requires a reference image to guide style (cannot work from scratch)
Effectiveness depends on the underlying AI model's ability to follow complex instructions
Only outputs matplotlib figures (not other plotting libraries)
May require trial and error to achieve optimal results with different reference images

Best For

Creating figures for machine learning and computer science research papersReplicating the visual style of existing published figures on new dataGenerating camera-ready figures with consistent aesthetics across multiple plotsAutomating figure generation for reproducible research

FAQ

What inputs does the tool require?

A reference image (reference.png) and user data in any parseable form such as CSV, TSV, markdown table, pasted table, or dirty terminal text. An optional max_iters parameter (default 6) controls the number of refinement iterations.

Does it copy the exact layout from the reference figure?

No. The reference image serves as a style anchor only (palette, spine treatment, gridline weight, marker shape, legend frame style, typographic voice, panel grid composition). Exact figsize, spacing, limits, font sizes, and annotation offsets are determined by the user's data.

Is this tool free?

Yes, the prompt is open-source and freely available on GitHub as part of the ai-boost/awesome-prompts repository. There are no usage fees.

What happens if the reference image has low resolution or ambiguous details?

The prompt includes an aesthetic convention library (L2) that defines default behaviors for properties that cannot be reliably estimated from the reference. When L1 (reference) and L2 conflict, L1 wins for PIL-reliable properties; L2 wins for properties that are unreliable to estimate.