diff --git a/examples/meson.build b/examples/meson.build
index 292ab053..a29a3310 100644
--- a/examples/meson.build
+++ b/examples/meson.build
@@ -26,5 +26,5 @@ executable(
 executable(
 	'screenshot',
 	'screenshot.c',
-	dependencies: [wayland_client, wlr_protos, dependency('cairo')],
+	dependencies: [wayland_client, wlr_protos],
 )
diff --git a/examples/screenshot.c b/examples/screenshot.c
index f70b9b56..253638c9 100644
--- a/examples/screenshot.c
+++ b/examples/screenshot.c
@@ -33,7 +33,6 @@
 #include <wayland-client.h>
 #include <limits.h>
 #include <sys/param.h>
-#include <cairo.h>
 #include <screenshooter-client-protocol.h>
 #include "../backend/wayland/os-compatibility.c"
 
@@ -174,11 +173,20 @@ create_shm_buffer(int width, int height, void **data_out)
 	return buffer;
 }
 
+static void argb_to_rgba(uint32_t *data, size_t height, size_t stride) {
+	size_t n = height*stride/4;
+	for (size_t i = 0; i < n; ++i) {
+		uint32_t v = data[i];
+		uint32_t rgb = v & 0x00ffffff;
+		uint32_t a = (v & 0xff000000) >> 24;
+		data[i] = (rgb << 8) | a;
+	}
+}
+
 static void
 write_png(int width, int height)
 {
 	int output_stride, buffer_stride, i;
-	cairo_surface_t *surface;
 	void *data, *d, *s;
 	struct screenshooter_output *output, *next;
 
@@ -203,11 +211,13 @@ write_png(int width, int height)
 		free(output);
 	}
 
-	surface = cairo_image_surface_create_for_data(data,
-						      CAIRO_FORMAT_RGB24,
-						      width, height, buffer_stride);
-	cairo_surface_write_to_png(surface, "wayland-screenshot.png");
-	cairo_surface_destroy(surface);
+	argb_to_rgba(data, height, buffer_stride);
+
+	// TODO: call convert
+	FILE *f = fopen("wayland-screenshot", "w");
+	fwrite(data, buffer_stride * height, 1, f);
+	fclose(f);
+
 	free(data);
 }
 
diff --git a/include/wlr/types/wlr_output.h b/include/wlr/types/wlr_output.h
index 7fab1cf6..ab3be5c7 100644
--- a/include/wlr/types/wlr_output.h
+++ b/include/wlr/types/wlr_output.h
@@ -71,6 +71,10 @@ void wlr_output_swap_buffers(struct wlr_output *output);
 void wlr_output_set_gamma(struct wlr_output *output,
 	uint16_t size, uint16_t *r, uint16_t *g, uint16_t *b);
 uint16_t wlr_output_get_gamma_size(struct wlr_output *output);
+
+/**
+ * Reads all pixels from the output and stores them as ARGB.
+ */
 void wlr_output_read_pixels(struct wlr_output *output, void *out_data);
 
 #endif
diff --git a/types/wlr_output.c b/types/wlr_output.c
index 85357d2a..1db63151 100644
--- a/types/wlr_output.c
+++ b/types/wlr_output.c
@@ -8,7 +8,6 @@
 #include <wlr/util/list.h>
 #include <wlr/util/log.h>
 #include <GLES2/gl2.h>
-#include <GLES2/gl2ext.h>
 #include <wlr/render/matrix.h>
 #include <wlr/render/gles2.h>
 #include <wlr/render.h>
@@ -252,9 +251,19 @@ uint16_t wlr_output_get_gamma_size(struct wlr_output *output) {
 	return output->impl->get_gamma_size(output);
 }
 
-void wlr_output_read_pixels(struct wlr_output *output, void *out_data) {
-	// TODO: is wlr_output_make_current required?
-	wlr_output_make_current(output);
-	glReadPixels(0, 0, output->width, output->height, GL_BGRA_EXT,
-		GL_UNSIGNED_BYTE, out_data);
+static void rgba_to_argb(uint32_t *data, size_t height, size_t stride) {
+	size_t n = height*stride/4;
+	for (size_t i = 0; i < n; ++i) {
+		uint32_t v = data[i];
+		uint32_t rgb = (v & 0xffffff00) >> 8;
+		uint32_t a = v & 0x000000ff;
+		data[i] = rgb | (a << 24);
+	}
+}
+
+void wlr_output_read_pixels(struct wlr_output *output, void *out_data) {
+	wlr_output_make_current(output);
+	glReadPixels(0, 0, output->width, output->height, GL_RGBA, GL_UNSIGNED_BYTE,
+		out_data);
+	rgba_to_argb(out_data, output->height, output->width*4);
 }