diff --git a/player/screenshot.c b/player/screenshot.c
index 7d1c743e40..2a1aaf6746 100644
--- a/player/screenshot.c
+++ b/player/screenshot.c
@@ -326,37 +326,38 @@ static void screenshot_save(struct MPContext *mpctx, struct mp_image *image)
 static struct mp_image *screenshot_get(struct MPContext *mpctx, int mode)
 {
     struct mp_image *image = NULL;
-    if (mpctx->video_out && mpctx->video_out->config_ok) {
-        if (mode == MODE_SUBTITLES && osd_get_render_subs_in_filter(mpctx->osd))
-            mode = 0;
+    if (mode == MODE_SUBTITLES && osd_get_render_subs_in_filter(mpctx->osd))
+        mode = 0;
 
-        struct voctrl_screenshot_args args =
-                            { .full_window = (mode == MODE_FULL_WINDOW) };
+    // vf_screenshot
+    if (mpctx->d_video && mpctx->d_video->vfilter)
+        vf_control_any(mpctx->d_video->vfilter, VFCTRL_SCREENSHOT, &image);
 
-        if (mpctx->d_video && mpctx->d_video->vfilter)
-            vf_control_any(mpctx->d_video->vfilter, VFCTRL_SCREENSHOT, &args);
+    if (!image && mpctx->video_out && mpctx->video_out->config_ok) {
+        vo_wait_frame(mpctx->video_out); // important for each-frame mode
 
-        if (!args.out_image) {
-            vo_wait_frame(mpctx->video_out); // important for each-frame mode
-            vo_control(mpctx->video_out, VOCTRL_SCREENSHOT, &args);
-        }
-
-        image = args.out_image;
-        if (image) {
-            if (mpctx->d_video && mpctx->d_video->hwdec_info) {
-                struct mp_hwdec_ctx *ctx = mpctx->d_video->hwdec_info->hwctx;
-                struct mp_image *nimage = NULL;
-                if (ctx && ctx->download_image)
-                    nimage = ctx->download_image(ctx, image, NULL);
-                if (nimage) {
-                    talloc_free(image);
-                    image = nimage;
-                }
-            }
-            if (mode == MODE_SUBTITLES && !args.has_osd)
-                add_subs(mpctx, image);
+        if (mode != MODE_FULL_WINDOW)
+            vo_control(mpctx->video_out, VOCTRL_SCREENSHOT, &image);
+        if (!image) {
+            vo_control(mpctx->video_out, VOCTRL_SCREENSHOT_WIN, &image);
+            mode = MODE_FULL_WINDOW;
         }
     }
+
+    if (image && mpctx->d_video && mpctx->d_video->hwdec_info) {
+        struct mp_hwdec_ctx *ctx = mpctx->d_video->hwdec_info->hwctx;
+        struct mp_image *nimage = NULL;
+        if (ctx && ctx->download_image)
+            nimage = ctx->download_image(ctx, image, NULL);
+        if (nimage) {
+            talloc_free(image);
+            image = nimage;
+        }
+    }
+
+    if (image && mode == MODE_SUBTITLES)
+        add_subs(mpctx, image);
+
     return image;
 }
 
diff --git a/video/filter/vf.h b/video/filter/vf.h
index 07f38d680b..3362320eb8 100644
--- a/video/filter/vf.h
+++ b/video/filter/vf.h
@@ -144,7 +144,7 @@ enum vf_ctrl {
     VFCTRL_SEEK_RESET = 1,   // reset on picture and PTS discontinuities
     VFCTRL_SET_EQUALIZER,    // set color options (brightness,contrast etc)
     VFCTRL_GET_EQUALIZER,    // get color options (brightness,contrast etc)
-    VFCTRL_SCREENSHOT,       // Take screenshot, arg is voctrl_screenshot_args
+    VFCTRL_SCREENSHOT,       // Take screenshot, arg is mp_image**
     VFCTRL_INIT_OSD,         // Filter OSD renderer present?
     VFCTRL_SET_DEINTERLACE,  // Set deinterlacing status
     VFCTRL_GET_DEINTERLACE,  // Get deinterlacing status
diff --git a/video/filter/vf_screenshot.c b/video/filter/vf_screenshot.c
index d60271acc4..562a1d03f5 100644
--- a/video/filter/vf_screenshot.c
+++ b/video/filter/vf_screenshot.c
@@ -46,8 +46,7 @@ static struct mp_image *filter(struct vf_instance *vf, struct mp_image *mpi)
 static int control (vf_instance_t *vf, int request, void *data)
 {
     if (request == VFCTRL_SCREENSHOT && vf->priv->current) {
-        struct voctrl_screenshot_args *args = data;
-        args->out_image = mp_image_new_ref(vf->priv->current);
+        *(struct mp_image **)data = mp_image_new_ref(vf->priv->current);
         return CONTROL_TRUE;
     }
     return CONTROL_UNKNOWN;
diff --git a/video/out/vo.h b/video/out/vo.h
index 281152352d..9ae489e2cd 100644
--- a/video/out/vo.h
+++ b/video/out/vo.h
@@ -97,7 +97,10 @@ enum mp_voctrl {
     // imgfmt/w/h/d_w/d_h can be omitted for convenience.
     VOCTRL_GET_COLORSPACE,              // struct mp_image_params*
 
-    VOCTRL_SCREENSHOT,                  // struct voctrl_screenshot_args*
+    // Retrieve original image.
+    VOCTRL_SCREENSHOT,                  // struct mp_image**
+    // Retrieve window contents.
+    VOCTRL_SCREENSHOT_WIN,              // struct mp_image**
 
     VOCTRL_SET_COMMAND_LINE,            // char**
 
@@ -120,25 +123,6 @@ struct voctrl_get_equalizer_args {
     int *valueptr;
 };
 
-// VOCTRL_SCREENSHOT
-struct voctrl_screenshot_args {
-    // 0: Save image of the currently displayed video frame, in original
-    //    resolution.
-    // 1: Save full screenshot of the window. Should contain OSD, EOSD, and the
-    //    scaled video.
-    // The value of this variable can be ignored if only a single method is
-    // implemented.
-    int full_window;
-    // Will be set to a newly allocated image, that contains the screenshot.
-    // The caller has to free the image with talloc_free().
-    // It is not specified whether the image data is a copy or references the
-    // image data directly.
-    // Can be NULL on failure.
-    struct mp_image *out_image;
-    // Whether the VO rendered OSD/subtitles into out_image
-    bool has_osd;
-};
-
 // VOCTRL_GET_WIN_STATE
 #define VO_WIN_STATE_MINIMIZED 1
 
diff --git a/video/out/vo_direct3d.c b/video/out/vo_direct3d.c
index 293eff55fe..5b4fb7856a 100644
--- a/video/out/vo_direct3d.c
+++ b/video/out/vo_direct3d.c
@@ -1260,14 +1260,12 @@ static int control(struct vo *vo, uint32_t request, void *data)
         return VO_TRUE;
     case VOCTRL_GET_PANSCAN:
         return VO_TRUE;
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
-        if (args->full_window)
-            args->out_image = get_window_screenshot(priv);
-        else
-            args->out_image = get_screenshot(priv);
-        return !!args->out_image;
-    }
+    case VOCTRL_SCREENSHOT:
+        *(struct mp_image **)data = get_screenshot(priv);
+        return VO_TRUE;
+    case VOCTRL_SCREENSHOT_WIN:
+        *(struct mp_image **)data = get_window_screenshot(priv);
+        return VO_TRUE;
     }
 
     int events = 0;
diff --git a/video/out/vo_opengl.c b/video/out/vo_opengl.c
index 6a2a48fb47..618ed90aff 100644
--- a/video/out/vo_opengl.c
+++ b/video/out/vo_opengl.c
@@ -348,13 +348,13 @@ static int control(struct vo *vo, uint32_t request, void *data)
         gl_video_get_colorspace(p->renderer, data);
         mpgl_unlock(p->glctx);
         return VO_TRUE;
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
+    case VOCTRL_SCREENSHOT_WIN:
+    case VOCTRL_SCREENSHOT:
+    {
         mpgl_lock(p->glctx);
-        if (args->full_window)
-            args->out_image = glGetWindowScreenshot(p->gl);
-        else
-            args->out_image = gl_video_download_image(p->renderer);
+        *(struct mp_image **)data = request == VOCTRL_SCREENSHOT
+                                  ? gl_video_download_image(p->renderer)
+                                  : glGetWindowScreenshot(p->gl);
         mpgl_unlock(p->glctx);
         return true;
     }
diff --git a/video/out/vo_opengl_cb.c b/video/out/vo_opengl_cb.c
index 091a376dfa..a974305b85 100644
--- a/video/out/vo_opengl_cb.c
+++ b/video/out/vo_opengl_cb.c
@@ -489,9 +489,8 @@ static int control(struct vo *vo, uint32_t request, void *data)
         pthread_mutex_unlock(&p->ctx->lock);
         return VO_TRUE;
     case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
         pthread_mutex_lock(&p->ctx->lock);
-        args->out_image = mp_image_new_ref(p->ctx->displayed_frame);
+        *(struct mp_image **)data = mp_image_new_ref(p->ctx->displayed_frame);
         pthread_mutex_unlock(&p->ctx->lock);
         return VO_TRUE;
     }
diff --git a/video/out/vo_sdl.c b/video/out/vo_sdl.c
index bc5ddda5c6..9b7605643a 100644
--- a/video/out/vo_sdl.c
+++ b/video/out/vo_sdl.c
@@ -1018,14 +1018,12 @@ static int control(struct vo *vo, uint32_t request, void *data)
         struct voctrl_get_equalizer_args *args = data;
         return get_eq(vo, args->name, args->valueptr);
     }
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
-        if (args->full_window)
-            args->out_image = get_window_screenshot(vo);
-        else
-            args->out_image = get_screenshot(vo);
+    case VOCTRL_SCREENSHOT:
+        *(struct mp_image **)data = get_screenshot(vo);
+        return true;
+    case VOCTRL_SCREENSHOT_WIN:
+        *(struct mp_image **)data = get_window_screenshot(vo);
         return true;
-    }
     case VOCTRL_SET_CURSOR_VISIBILITY:
         SDL_ShowCursor(*(bool *)data);
         return true;
diff --git a/video/out/vo_vaapi.c b/video/out/vo_vaapi.c
index 06c7b3c9f4..1c4b016e6b 100644
--- a/video/out/vo_vaapi.c
+++ b/video/out/vo_vaapi.c
@@ -541,11 +541,9 @@ static int control(struct vo *vo, uint32_t request, void *data)
         p->output_surface = p->visible_surface;
         draw_osd(vo);
         return true;
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
-        args->out_image = get_screenshot(p);
+    case VOCTRL_SCREENSHOT:
+        *(struct mp_image **)data = get_screenshot(p);
         return true;
-    }
     case VOCTRL_GET_PANSCAN:
         return VO_TRUE;
     case VOCTRL_SET_PANSCAN:
diff --git a/video/out/vo_vdpau.c b/video/out/vo_vdpau.c
index 7b98b84243..1a1287c1d7 100644
--- a/video/out/vo_vdpau.c
+++ b/video/out/vo_vdpau.c
@@ -1048,14 +1048,15 @@ static int control(struct vo *vo, uint32_t request, void *data)
     case VOCTRL_RESET:
         forget_frames(vo, true);
         return true;
-    case VOCTRL_SCREENSHOT: {
+    case VOCTRL_SCREENSHOT_WIN:
+    case VOCTRL_SCREENSHOT:
+    {
         if (!status_ok(vo))
             return false;
-        struct voctrl_screenshot_args *args = data;
-        if (args->full_window) {
-            args->out_image = get_window_screenshot(vo);
+        if (request == VOCTRL_SCREENSHOT_WIN) {
+            *(struct mp_image **)data = get_window_screenshot(vo);
         } else {
-            args->out_image =
+            *(struct mp_image **)data =
                 vc->current_image ? mp_image_new_ref(vc->current_image) : NULL;
         }
         return true;
diff --git a/video/out/vo_wayland.c b/video/out/vo_wayland.c
index 770142ae9d..ed23f8af3d 100644
--- a/video/out/vo_wayland.c
+++ b/video/out/vo_wayland.c
@@ -690,11 +690,8 @@ static int control(struct vo *vo, uint32_t request, void *data)
     case VOCTRL_REDRAW_FRAME:
         return redraw_frame(p);
     case VOCTRL_SCREENSHOT:
-    {
-        struct voctrl_screenshot_args *args = data;
-        args->out_image = get_screenshot(p);
-        return VO_TRUE;
-    }
+        *(struct mp_image **)data = get_screenshot(p);
+        return true;
     case VOCTRL_GET_RECENT_FLIP_TIME:
     {
         *(int64_t*) data = p->recent_flip_time;
diff --git a/video/out/vo_x11.c b/video/out/vo_x11.c
index 13fa0a3c9f..32d9b731fd 100644
--- a/video/out/vo_x11.c
+++ b/video/out/vo_x11.c
@@ -619,12 +619,10 @@ static int control(struct vo *vo, uint32_t request, void *data)
     case VOCTRL_REDRAW_FRAME:
         draw_image(vo, p->original_image);
         return true;
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
-        args->out_image = get_screenshot(vo);
+    case VOCTRL_SCREENSHOT:
+        *(struct mp_image **)data = get_screenshot(vo);
         return true;
     }
-    }
 
     int events = 0;
     int r = vo_x11_control(vo, &events, request, data);
diff --git a/video/out/vo_xv.c b/video/out/vo_xv.c
index 202ba13735..e506fc9359 100644
--- a/video/out/vo_xv.c
+++ b/video/out/vo_xv.c
@@ -839,12 +839,10 @@ static int control(struct vo *vo, uint32_t request, void *data)
     case VOCTRL_REDRAW_FRAME:
         draw_image(vo, ctx->original_image);
         return true;
-    case VOCTRL_SCREENSHOT: {
-        struct voctrl_screenshot_args *args = data;
-        args->out_image = get_screenshot(vo);
+    case VOCTRL_SCREENSHOT:
+        *(struct mp_image **)data = get_screenshot(vo);
         return true;
     }
-    }
     int events = 0;
     int r = vo_x11_control(vo, &events, request, data);
     if (events & (VO_EVENT_EXPOSE | VO_EVENT_RESIZE))