diff --git a/paperforge/worker/ocr.py b/paperforge/worker/ocr.py
index e5e680d5..e483f7f0 100644
--- a/paperforge/worker/ocr.py
+++ b/paperforge/worker/ocr.py
@@ -1779,6 +1779,20 @@ def postprocess_ocr_result(vault: Path, key: str, all_results: list[dict]) -> tu
         table_inventory,
     )
 
+    # --- Phase 2: object artifacts ---
+    from paperforge.worker.ocr_objects import extract_and_write_objects
+
+    ocr_asset_root = ocr_root / "assets"
+    ocr_render_root = ocr_root / "render"
+
+    extract_and_write_objects(
+        pdf_path=source_pdf_path,
+        figure_inventory=figure_inventory,
+        table_inventory=table_inventory,
+        asset_root=ocr_asset_root,
+        render_root=ocr_render_root,
+    )
+
     # Update meta.json with version payloads
     ocr_model = meta.get("ocr_model", meta.get("ocr_provider", "PaddleOCR"))
     version_payload = build_version_payload(
diff --git a/paperforge/worker/ocr_objects.py b/paperforge/worker/ocr_objects.py
new file mode 100644
index 00000000..532c83b3
--- /dev/null
+++ b/paperforge/worker/ocr_objects.py
@@ -0,0 +1,229 @@
+from __future__ import annotations
+
+from pathlib import Path
+from typing import Any
+
+
+
+def render_figure_object_markdown(figure: dict[str, Any]) -> str:
+    """Render object markdown for a figure entry.
+
+    Produces Obsidian-flavored markdown with a wikilink embed to the
+    cropped asset image and the caption text.
+    """
+    caption = figure.get("caption", "")
+    image_relpath = figure.get("image_relpath", "")
+
+    label = ""
+    if caption:
+        label = caption.split(".")[0].strip()
+    if not label:
+        figure_id = figure.get("figure_id", "")
+        if figure_id.startswith("orphan_"):
+            label = "Orphan Media"
+        else:
+            label = f"Figure {figure_id}"
+
+    parts = [f"# {label}", "", f"![[../{image_relpath}]]", ""]
+    if caption:
+        parts.append(caption)
+    if figure.get("page"):
+        parts.append("")
+        parts.append(f"*Page {figure['page']}*")
+    parts.append("")
+    parts.append("---")
+
+    return "\n".join(parts)
+
+
+def render_table_object_markdown(table: dict[str, Any]) -> str:
+    """Render object markdown for a table entry."""
+    caption = table.get("caption", "")
+    image_relpath = table.get("image_relpath", "")
+
+    label = ""
+    if caption:
+        label = caption.split(".")[0].strip()
+    if not label:
+        label = f"Table {table.get('table_id', 'unknown')}"
+
+    parts = [f"# {label}", "", f"![[../{image_relpath}]]", ""]
+    if caption:
+        parts.append(caption)
+    if table.get("page"):
+        parts.append("")
+        parts.append(f"*Page {table['page']}*")
+    parts.append("")
+    parts.append("---")
+
+    return "\n".join(parts)
+
+
+def _write_object_markdown(md: str, dst: Path) -> None:
+    dst.parent.mkdir(parents=True, exist_ok=True)
+    dst.write_text(md.strip() + "\n", encoding="utf-8")
+
+
+def _crop_asset_from_pdf(
+    pdf_path: Path,
+    page_num: int,
+    bbox: list[float],
+    dst: Path,
+) -> bool:
+    """Crop a region from a PDF page and save as JPEG.
+
+    Args:
+        pdf_path: Path to the PDF file.
+        page_num: 1-based page number.
+        bbox: [x1, y1, x2, y2] bounding box in PDF coordinates.
+        dst: Output path for the JPEG file.
+
+    Returns:
+        True if crop succeeded, False otherwise.
+    """
+    try:
+        import fitz
+    except ImportError:
+        return False
+
+    if not pdf_path.exists():
+        return False
+
+    try:
+        doc = fitz.open(str(pdf_path))
+        page = doc[page_num - 1]
+        rect = fitz.Rect(*bbox)
+        pix = page.get_pixmap(matrix=fitz.Matrix(2, 2), clip=rect)
+        dst.parent.mkdir(parents=True, exist_ok=True)
+        pix.save(str(dst))
+        doc.close()
+        return True
+    except Exception:
+        return False
+
+
+def extract_and_write_objects(
+    pdf_path: Path | None,
+    figure_inventory: dict[str, Any],
+    table_inventory: dict[str, Any],
+    asset_root: Path,
+    render_root: Path,
+) -> None:
+    """Extract figure/table asset crops from PDF and write object markdown.
+
+    Args:
+        pdf_path: Path to the source PDF (may be None).
+        figure_inventory: The figure inventory dict.
+        table_inventory: The table inventory dict.
+        asset_root: Root directory for assets (e.g., <ocr_root>/assets/).
+        render_root: Root directory for render objects (e.g., <ocr_root>/render/).
+    """
+    figures_asset_dir = asset_root / "figures"
+    tables_asset_dir = asset_root / "tables"
+    orphans_asset_dir = asset_root / "orphans"
+    figures_render_dir = render_root / "figures"
+    tables_render_dir = render_root / "tables"
+
+    for d in (figures_asset_dir, tables_asset_dir, orphans_asset_dir,
+              figures_render_dir, tables_render_dir):
+        d.mkdir(parents=True, exist_ok=True)
+
+    # Process matched figures
+    for i, match in enumerate(figure_inventory.get("matched_figures", [])):
+        fig_id = f"figure_{i + 1:03d}"
+        caption_text = match.get("text", "")
+        page = match.get("page", 0)
+        asset_path_rel = f"assets/figures/{fig_id}.jpg"
+        asset_path_abs = figures_asset_dir / f"{fig_id}.jpg"
+
+        was_cropped = False
+        for asset_info in match.get("matched_assets", []):
+            bbox = asset_info.get("bbox", [0, 0, 0, 0])
+            if pdf_path and bbox and all(v > 0 for v in bbox):
+                if _crop_asset_from_pdf(pdf_path, page, bbox, asset_path_abs):
+                    was_cropped = True
+                    break
+
+        if not was_cropped:
+            for asset in figure_inventory.get("unmatched_assets", []):
+                bbox = asset.get("bbox", [0, 0, 0, 0])
+                asset_page = asset.get("page", 0)
+                if pdf_path and bbox and all(v > 0 for v in bbox):
+                    if _crop_asset_from_pdf(pdf_path, asset_page, bbox, asset_path_abs):
+                        was_cropped = True
+                        break
+
+        md = render_figure_object_markdown({
+            "figure_id": fig_id,
+            "page": page,
+            "caption": caption_text,
+            "image_relpath": asset_path_rel,
+            "confidence": match.get("confidence", 0.5),
+        })
+        _write_object_markdown(md, figures_render_dir / f"{fig_id}.md")
+
+    # Process unmatched assets as orphans
+    orphan_count = 0
+    for asset in figure_inventory.get("unmatched_assets", []):
+        orphan_count += 1
+        orphan_id = f"orphan_{orphan_count:03d}"
+        bbox = asset.get("bbox", [0, 0, 0, 0])
+        page = asset.get("page", 0)
+        asset_path_rel = f"assets/orphans/{orphan_id}.jpg"
+        asset_path_abs = orphans_asset_dir / f"{orphan_id}.jpg"
+
+        was_cropped = False
+        if pdf_path and bbox and all(v > 0 for v in bbox):
+            was_cropped = _crop_asset_from_pdf(pdf_path, page, bbox, asset_path_abs)
+
+        md = render_figure_object_markdown({
+            "figure_id": orphan_id,
+            "page": page,
+            "caption": "",
+            "image_relpath": asset_path_rel,
+            "confidence": 0.3,
+        })
+        _write_object_markdown(md, figures_render_dir / f"{orphan_id}.md")
+
+    # Process tables
+    for i, table in enumerate(table_inventory.get("tables", [])):
+        tbl_id = f"table_{i + 1:03d}"
+        caption_text = table.get("caption_text", "")
+        page = table.get("page", 0)
+        asset_bbox = table.get("asset_bbox", [0, 0, 0, 0])
+        asset_path_rel = f"assets/tables/{tbl_id}.jpg"
+        asset_path_abs = tables_asset_dir / f"{tbl_id}.jpg"
+
+        was_cropped = False
+        if table.get("has_asset") and pdf_path and asset_bbox and all(v > 0 for v in asset_bbox):
+            was_cropped = _crop_asset_from_pdf(pdf_path, page, asset_bbox, asset_path_abs)
+
+        md = render_table_object_markdown({
+            "table_id": tbl_id,
+            "page": page,
+            "caption": caption_text,
+            "image_relpath": asset_path_rel,
+            "confidence": 0.85 if was_cropped else 0.4,
+        })
+        _write_object_markdown(md, tables_render_dir / f"{tbl_id}.md")
+
+    # Process unmatched table assets as orphans
+    for asset in table_inventory.get("unmatched_assets", []):
+        orphan_count += 1
+        orphan_id = f"orphan_{orphan_count:03d}"
+        bbox = asset.get("bbox", [0, 0, 0, 0])
+        page = asset.get("page", 0)
+        asset_path_rel = f"assets/orphans/{orphan_id}.jpg"
+        asset_path_abs = orphans_asset_dir / f"{orphan_id}.jpg"
+
+        if pdf_path and bbox and all(v > 0 for v in bbox):
+            _crop_asset_from_pdf(pdf_path, page, bbox, asset_path_abs)
+
+        md = render_figure_object_markdown({
+            "figure_id": orphan_id,
+            "page": page,
+            "caption": "",
+            "image_relpath": asset_path_rel,
+            "confidence": 0.3,
+        })
+        _write_object_markdown(md, figures_render_dir / f"{orphan_id}.md")
diff --git a/paperforge/worker/ocr_tables.py b/paperforge/worker/ocr_tables.py
index 3f44d05c..d610a78d 100644
--- a/paperforge/worker/ocr_tables.py
+++ b/paperforge/worker/ocr_tables.py
@@ -6,7 +6,6 @@ from typing import Any
 
 from paperforge.core.io import write_json
 
-
 _TABLE_PREFIX_PATTERN = re.compile(
     r"^(?:Table|Supplementary\s+Table|Extended\s+Data\s+Table)\s+(\d+(?:\.\d+)?)",
     flags=re.IGNORECASE,
@@ -60,24 +59,22 @@ def build_table_inventory(structured_blocks: list[dict]) -> dict[str, Any]:
             if matched_asset:
                 break
 
-        tables.append({
-            "caption_block_id": caption.get("block_id", ""),
-            "page": caption_page,
-            "caption_text": caption_text,
-            "table_number": table_num,
-            "asset_block_id": matched_asset.get("block_id", "") if matched_asset else "",
-            "asset_bbox": matched_asset.get("bbox", [0, 0, 0, 0]) if matched_asset else [],
-            "assistive_text": (matched_asset.get("text", "") or "") if matched_asset else "",
-            "truth_source": "image",
-            "has_asset": matched_asset is not None,
-        })
+        tables.append(
+            {
+                "caption_block_id": caption.get("block_id", ""),
+                "page": caption_page,
+                "caption_text": caption_text,
+                "table_number": table_num,
+                "asset_block_id": matched_asset.get("block_id", "") if matched_asset else "",
+                "asset_bbox": matched_asset.get("bbox", [0, 0, 0, 0]) if matched_asset else [],
+                "assistive_text": (matched_asset.get("text", "") or "") if matched_asset else "",
+                "truth_source": "image",
+                "has_asset": matched_asset is not None,
+            }
+        )
 
     for caption in captions:
-        if not any(
-            t["caption_block_id"] == caption.get("block_id", "")
-            for t in tables
-            if t["has_asset"]
-        ):
+        if not any(t["caption_block_id"] == caption.get("block_id", "") for t in tables if t["has_asset"]):
             unmatched_captions.append(caption)
 
     for i, asset in enumerate(assets):
diff --git a/tests/test_ocr.py b/tests/test_ocr.py
index 3205d478..d3f9b676 100644
--- a/tests/test_ocr.py
+++ b/tests/test_ocr.py
@@ -379,3 +379,34 @@ def test_postprocess_writes_table_inventory(tmp_path: Path) -> None:
 
     # Phase 2: table_inventory.json does not exist yet -- will fail
     assert (ocr_dir / "structure" / "table_inventory.json").exists()
+
+
+def test_postprocess_creates_object_directories(tmp_path: Path) -> None:
+    """Verify postprocess creates object directories even with empty results."""
+    import json
+
+    from paperforge.worker.ocr import postprocess_ocr_result
+
+    vault = tmp_path / "vault"
+    vault.mkdir()
+    (vault / "paperforge.json").write_text(
+        json.dumps({"vault_config": {"system_dir": "System", "resources_dir": "Resources"}}),
+        encoding="utf-8",
+    )
+    ocr_root = vault / "System" / "PaperForge" / "ocr"
+    ocr_root.mkdir(parents=True)
+    ocr_dir = ocr_root / "OBJ001"
+    ocr_dir.mkdir()
+    (ocr_dir / "meta.json").write_text(
+        '{"zotero_key":"OBJ001","ocr_status":"done","ocr_model":"PaddleOCR"}',
+        encoding="utf-8",
+    )
+
+    _, _, _, _ = postprocess_ocr_result(vault, "OBJ001", [])
+
+    # Phase 2: object artifact directories are created during postprocess
+    assert (ocr_dir / "render" / "figures").exists()
+    assert (ocr_dir / "render" / "tables").exists()
+    assert (ocr_dir / "assets" / "figures").exists()
+    assert (ocr_dir / "assets" / "tables").exists()
+    assert (ocr_dir / "assets" / "orphans").exists()
diff --git a/tests/test_ocr_objects.py b/tests/test_ocr_objects.py
new file mode 100644
index 00000000..526f36e6
--- /dev/null
+++ b/tests/test_ocr_objects.py
@@ -0,0 +1,49 @@
+from __future__ import annotations
+
+
+
+def test_figure_object_markdown_links_image_and_legend() -> None:
+    from paperforge.worker.ocr_objects import render_figure_object_markdown
+
+    md = render_figure_object_markdown({
+        "figure_id": "figure_001",
+        "page": 4,
+        "caption": "Figure 1. Example.",
+        "image_relpath": "assets/figures/figure_001.jpg",
+        "confidence": 0.91,
+    })
+
+    assert "# Figure 1" in md
+    assert "![[../assets/figures/figure_001.jpg]]" in md
+    assert "Figure 1. Example." in md
+
+
+def test_table_object_markdown_includes_image_and_caption() -> None:
+    from paperforge.worker.ocr_objects import render_table_object_markdown
+
+    md = render_table_object_markdown({
+        "table_id": "table_001",
+        "page": 5,
+        "caption": "Table 1. Results.",
+        "image_relpath": "assets/tables/table_001.jpg",
+        "confidence": 0.88,
+    })
+
+    assert "# Table 1" in md
+    assert "![[../assets/tables/table_001.jpg]]" in md
+    assert "Table 1. Results." in md
+
+
+def test_orphan_object_markdown() -> None:
+    from paperforge.worker.ocr_objects import render_figure_object_markdown
+
+    md = render_figure_object_markdown({
+        "figure_id": "orphan_001",
+        "page": 6,
+        "caption": "",
+        "image_relpath": "assets/orphans/orphan_001.jpg",
+        "confidence": 0.3,
+    })
+
+    assert "# Orphan Media" in md
+    assert "![[../assets/orphans/orphan_001.jpg]]" in md