diff --git a/libavutil/hwcontext_vulkan.c b/libavutil/hwcontext_vulkan.c
index 47feff43dc1443b3e840ae50b7b56aadbedb48cf..9374b8663286a844f763c8f30cd4fb4fd7db0890 100644
--- a/libavutil/hwcontext_vulkan.c
+++ b/libavutil/hwcontext_vulkan.c
@@ -1703,7 +1703,8 @@ static int vulkan_frames_init(AVHWFramesContext *hwfc)
     hwctx->tiling = hwctx->tiling ? hwctx->tiling : p->use_linear_images ?
                     VK_IMAGE_TILING_LINEAR : VK_IMAGE_TILING_OPTIMAL;
 
-    hwctx->usage |= DEFAULT_USAGE_FLAGS;
+    if (!hwctx->usage)
+        hwctx->usage = DEFAULT_USAGE_FLAGS;
 
     err = create_exec_ctx(hwfc->device_ctx, &fp->conv_ctx,
                           dev_hwctx->queue_family_comp_index,
diff --git a/libavutil/hwcontext_vulkan.h b/libavutil/hwcontext_vulkan.h
index 656411e0a8275bbcc23f1219492cbe6e01376263..0ac34658b2a19ab1c43f4ad55289821f08c3d914 100644
--- a/libavutil/hwcontext_vulkan.h
+++ b/libavutil/hwcontext_vulkan.h
@@ -112,8 +112,8 @@ typedef struct AVVulkanFramesContext {
      */
     VkImageTiling tiling;
     /**
-     * Defines extra usage of output frames. This is bitwise OR'd with the
-     * standard usage flags (SAMPLED, STORAGE, TRANSFER_SRC and TRANSFER_DST).
+     * Defines extra usage of output frames. If left as 0, the following bits
+     * are set: TRANSFER_SRC, TRANSFER_DST. SAMPLED and STORAGE.
      */
     VkImageUsageFlagBits usage;
     /**