Figure Composer

Three-mode figure workflow: compose (from scratch), edit (modify existing), clean (PDF label removal).

Quick Dispatch

Mode 1: Compose (from scratch)

Phase 1: Collect & confirm layout

Ask the user for input files, rows, cols. Draw an ASCII diagram:

Proposed: 2×3 (5 files, 1 empty)
┌──────┬──────┬──────┐
│  a   │  b   │  c   │
├──────┼──────┼──────┤
│  d   │  e   │ (empty)│
└──────┴──────┴──────┘

Always confirm before proceeding.

Phase 2: Clean old labels (ask first)

Ask the user: "Detected X candidate labels. Remove them?" Default is yes, but the user can say "keep the original labels" to skip.

For PDF: python scripts/replace_pdf_label.py --pdf <file> --detect Then: python scripts/replace_pdf_label.py --pdf <file> --old "X" --new ""

For SVG, do inline search-and-remove. For PNG, overlay white rectangle with Pillow.

Phase 3: Compose

python scripts/compose.py \
  --files <file1> <file2> ... \
  --rows <N> --cols <N> \
  --labels "a,b,c,d,e" \
  --output <output.pptx>

• PDF → SVG via PyMuPDF (vector preserved)
• SVG/PNG embedded directly
• Labels added as native PPTX text boxes (Arial Bold, inside panel top-left)
• No Windows/PowerPoint dependency — pure Python (PyMuPDF + lxml + zipfile)

Phase 4: Verify

python scripts/manage_labels.py detect --pptx <output.pptx>
python scripts/replace_panel.py --pptx <output.pptx> --list

Tell user: right-click SVG → "Convert to Shape" for full editability.

Mode 2: Edit (modify existing PPTX)

Step 1: Inspect current state

python scripts/replace_panel.py --pptx <file.pptx> --list
python scripts/manage_labels.py detect --pptx <file.pptx>

Step 2a: Replace a panel

python scripts/replace_panel.py \
  --pptx <file.pptx> \
  --panel <media_name> \    # e.g. image1.svg
  --with <new_file.pdf> \   # PDF→SVG automatic
  --output <file.pptx>

Automatically creates .bak backup. Only page 1 of multi-page PDF is used.

Step 2b: Manage labels

# Add labels (mapped from SVG coords to slide position)
python scripts/manage_labels.py add \
  --pptx <file.pptx> \
  --labels "a,62,20 b,455,20 c,843,20" \
  --size 20 --bold \
  --panel-x <panel_emu_x> --panel-y <panel_emu_y> \
  --panel-w <panel_emu_w> --panel-h <panel_emu_h>

# Remove labels
python scripts/manage_labels.py remove --pptx <file.pptx> --labels "a,b,c"

# Modify label properties (font, size, bold)
python scripts/manage_labels.py modify \
  --pptx <file.pptx> --labels "a,b,c" --size 24 --font "Arial" --bold

Step 3: Verify

Re-run detect to confirm changes. To export a raster/vector copy afterwards, see the conversion pipeline in the r-figures-to-editable-pptx skill.

Mode 3: Clean (PDF label removal)

For sophisticated label replacement with font matching and border preservation:

# Detect labels in PDF
python scripts/replace_pdf_label.py --pdf <file.pdf> --detect

# Replace specific text (no-fill redaction preserves borders)
python scripts/replace_pdf_label.py --pdf <file.pdf> \
  --old "a Translocated species richness" --new ""

# Replace with different text at same position/size/font
python scripts/replace_pdf_label.py --pdf <file.pdf> \
  --old "a" --new "d" -o <output.pdf>

Key capabilities (absent from simpler tools):

• No-fill redaction: preserves border/axis lines passing behind the label
• Font matching: detects and matches Times/Arial/Helvetica/Courier
• Border integrity check: verifies surrounding drawing lines are intact
• Preview PNG: generates 150 DPI preview for visual confirmation

Scripts Reference

Dependencies

python -m pip install pymupdf pillow lxml python-pptx

• PyMuPDF (fitz) — PDF manipulation, PDF→SVG, label detection/redaction
• Pillow — PNG label cleaning
• lxml — PPTX XML manipulation
• python-pptx — PPTX base template creation (minimal use)
• Ghostscript (optional) — for preview PNG in label replacement
• No Windows/PowerPoint dependency for compose and label operations