diff --git a/docs/design/camera.md b/docs/design/camera.md
index 9baa44f0..601e8bb7 100644
--- a/docs/design/camera.md
+++ b/docs/design/camera.md
@@ -21,7 +21,9 @@ to grant it, and the grant visible for as long as it lasts.
   device made for it.
 - **Not covered:** the MIPI cameras of many Intel laptops since 2022, behind
   the IPU6, which need a camera stack in userspace (libcamera and Intel's
-  image processing), and infrared face-login cameras.
+  image processing); those behind the image processor of AMD's Ryzen AI Max
+  (HP's ZBook Ultra G1a), whose driver came in Linux 7.2, after 6.18; and
+  infrared face-login cameras.
 
 ## Who reaches that code, and when
 
@@ -115,9 +117,10 @@ itself waits for the tests on physical hardware.
 
 ## Not covered
 
-- MIPI cameras behind the IPU6 (below), and infrared cameras. Of several
-  cameras, a grant gives the zone the lowest numbered node that is its
-  device's first (`index` 0), which a camera's metadata node is not
+- MIPI cameras behind Intel's IPU6 and AMD's image processor (below), and
+  infrared cameras. Of several cameras, a grant gives the zone the lowest
+  numbered node that is its device's first (`index` 0), which a camera's
+  metadata node is not
   (`camera_in` in `camera.rs`). On a laptop whose infrared face-login camera
   is a USB device of its own, that node may be the infrared camera's, which
   gives a call no picture. Preferring a node that offers a colour format
@@ -159,3 +162,13 @@ it; a USB camera works on those laptops. The firmware is the small part of
 the cost. Where the image processing runs and what a grant hands over are
 the questions, and they wait for a machine with such a camera to answer
 them on.
+
+AMD's Ryzen AI Max has an image processor of its own, the ISP4, behind the
+MIPI camera of HP's ZBook Ultra G1a. Its firmware drives the sensor and
+makes the picture, and its driver gives one capture node of NV12 or YUYV
+frames, as a USB camera does, rather than a graph of raw ones, so it would
+fit a grant of one node. That driver (`VIDEO_AMD_ISP4_CAPTURE`) came in
+Linux 7.2, and Kryptik builds 6.18. It would also need amdgpu's half
+(`DRM_AMD_ISP`, off), platform cameras in the media filter (off), the
+processor's firmware (`amdgpu/isp_*`, which `firmware.list` leaves out),
+and kryptikd's grant loading it beside `uvcvideo`.
diff --git a/docs/user-guide.md b/docs/user-guide.md
index f8befcdf..ee798e63 100644
--- a/docs/user-guide.md
+++ b/docs/user-guide.md
@@ -423,7 +423,8 @@ as well. A window in full screen covers those windows, so the bar above it
 names the zone that may use the microphone (`MIC`) and the zone that has the
 camera (`CAM`), whichever zones they are. Built-in and USB cameras that speak the USB Video Class work, as
 most do; the cameras of recent Intel laptops that need Intel's image
-processing in userspace (MIPI cameras behind the IPU6) and infrared
+processing in userspace (MIPI cameras behind the IPU6), those behind the
+image processor of AMD's Ryzen AI Max (HP's ZBook Ultra G1a), and infrared
 face-login cameras do not. `vault` and the net zone never get the camera.
 
 Firefox in a zone is not told when the camera is given or taken back.