diff --git a/CHANGELOG.md b/CHANGELOG.md
index 2610305e..d98f0cf9 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -1,3 +1,8 @@
+## 1.6.13
+
+### Fixes
+- **Render filled PDF form fields**: `convert_pdf_to_image` now calls `init_forms()` so AcroForm/XFA field values (text typed into fillable fields) are painted into the rendered page image instead of being silently dropped.
+
 ## 1.6.12
 
 ### Fixes
diff --git a/sample-docs/form-field.pdf b/sample-docs/form-field.pdf
new file mode 100644
index 00000000..f4d45f22
Binary files /dev/null and b/sample-docs/form-field.pdf differ
diff --git a/test_unstructured_inference/inference/test_pdf_image_forms.py b/test_unstructured_inference/inference/test_pdf_image_forms.py
new file mode 100644
index 00000000..41be6939
--- /dev/null
+++ b/test_unstructured_inference/inference/test_pdf_image_forms.py
@@ -0,0 +1,52 @@
+from __future__ import annotations
+
+import numpy as np
+import pypdfium2 as pdfium
+
+from unstructured_inference.inference import pdf_image
+
+# sample-docs/form-field.pdf is a 1-page PDF with an empty content stream and a single
+# filled text form field. The field value ("FORMVALUE777") is drawn only by the widget
+# annotation's appearance stream, so it renders only when the form-fill environment is
+# initialized (init_forms). Geometry below mirrors the fixture's widget rectangle.
+FORM_PDF = "sample-docs/form-field.pdf"
+PAGE_WIDTH, PAGE_HEIGHT = 612, 792
+FIELD_RECT = (40, 700, 320, 724)  # x1, y1, x2, y2 in PDF user space (origin bottom-left)
+RENDER_DPI = 200
+
+
+def _field_region_dark_pixels(img) -> int:
+    """Count dark pixels inside the form field's rectangle in the rendered image."""
+    gray = img.convert("L")
+    scale_x = gray.width / PAGE_WIDTH
+    scale_y = gray.height / PAGE_HEIGHT
+    x0, y0, x1, y1 = FIELD_RECT
+    # PDF user space is bottom-up; image space is top-down.
+    box = (
+        int(x0 * scale_x),
+        int((PAGE_HEIGHT - y1) * scale_y),
+        int(x1 * scale_x),
+        int((PAGE_HEIGHT - y0) * scale_y),
+    )
+    crop = np.array(gray.crop(box))
+    return int(np.count_nonzero(crop < 128))
+
+
+def test_convert_pdf_to_image_renders_acroform_field_value():
+    """Filled form-field values are painted into the rendered page image."""
+    img = pdf_image.convert_pdf_to_image(filename=FORM_PDF, dpi=RENDER_DPI)[0]
+
+    assert _field_region_dark_pixels(img) > 100, "Expected the form field value to be rendered"
+
+
+def test_convert_pdf_to_image_drops_form_field_without_init_forms(monkeypatch):
+    """Control: without init_forms() the widget appearance is not painted.
+
+    Patching init_forms() to a no-op reproduces the pre-fix behavior and proves the
+    rendered field value in the test above comes specifically from initializing the
+    form-fill environment, not from the page content stream (which is empty here).
+    """
+    monkeypatch.setattr(pdfium.PdfDocument, "init_forms", lambda self, *a, **k: None)
+    img = pdf_image.convert_pdf_to_image(filename=FORM_PDF, dpi=RENDER_DPI)[0]
+
+    assert _field_region_dark_pixels(img) == 0, "Field should be blank without form init"
diff --git a/unstructured_inference/__version__.py b/unstructured_inference/__version__.py
index 1be475c7..9ae5e1d8 100644
--- a/unstructured_inference/__version__.py
+++ b/unstructured_inference/__version__.py
@@ -1 +1 @@
-__version__ = "1.6.12"  # pragma: no cover
+__version__ = "1.6.13"  # pragma: no cover
diff --git a/unstructured_inference/inference/pdf_image.py b/unstructured_inference/inference/pdf_image.py
index 2f16e18c..b6c359f0 100644
--- a/unstructured_inference/inference/pdf_image.py
+++ b/unstructured_inference/inference/pdf_image.py
@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import contextlib
 import math
 import os
 from functools import lru_cache
@@ -166,6 +167,13 @@ def _in_range(page_num: int) -> bool:
 
     with _pdfium_lock:
         pdf = pdfium.PdfDocument(filename or file, password=password)
+        # Initialize the form-fill environment so AcroForm/XFA field values
+        # (e.g. text typed into fillable fields) are painted into the rendered
+        # image. Without this, pdfium silently drops widget annotation content
+        # even though may_draw_forms defaults to True on page.render().
+        # Fall back to page rendering without form appearances when form env init fails.
+        with contextlib.suppress(pdfium.PdfiumError):
+            pdf.init_forms()
         n_pages = len(pdf)
 
         # Pre-scan page rotations so the (heavier) text-orientation pass only runs on the