diff --git a/libavcodec/mediacodec_surface.c b/libavcodec/mediacodec_surface.c
index 09a42295d2..ef41cdafa7 100644
--- a/libavcodec/mediacodec_surface.c
+++ b/libavcodec/mediacodec_surface.c
@@ -20,33 +20,59 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <android/native_window.h>
 #include <jni.h>
 
+#include "libavutil/mem.h"
 #include "ffjni.h"
 #include "mediacodec_surface.h"
 
-FFANativeWindow *ff_mediacodec_surface_ref(void *surface, void *log_ctx)
+FFANativeWindow *ff_mediacodec_surface_ref(void *surface, void *native_window, void *log_ctx)
 {
-    JNIEnv *env = NULL;
+    FFANativeWindow *ret;
 
-    env = ff_jni_get_env(log_ctx);
-    if (!env) {
+    ret = av_mallocz(sizeof(*ret));
+    if (!ret)
         return NULL;
+
+    if (surface) {
+        JNIEnv *env = NULL;
+
+        env = ff_jni_get_env(log_ctx);
+        if (env)
+            ret->surface = (*env)->NewGlobalRef(env, surface);
     }
 
-    return (*env)->NewGlobalRef(env, surface);
+    if (native_window) {
+        ANativeWindow_acquire(native_window);
+        ret->native_window = native_window;
+    }
+
+    if (!ret->surface && !ret->native_window) {
+        av_log(log_ctx, AV_LOG_ERROR, "Both surface and native_window are NULL\n");
+        av_freep(&ret);
+    }
+
+    return ret;
 }
 
 int ff_mediacodec_surface_unref(FFANativeWindow *window, void *log_ctx)
 {
-    JNIEnv *env = NULL;
+    if (!window)
+        return 0;
 
-    env = ff_jni_get_env(log_ctx);
-    if (!env) {
-        return AVERROR_EXTERNAL;
+    if (window->surface) {
+        JNIEnv *env = NULL;
+
+        env = ff_jni_get_env(log_ctx);
+        if (env)
+            (*env)->DeleteGlobalRef(env, window->surface);
     }
 
-    (*env)->DeleteGlobalRef(env, window);
+    if (window->native_window)
+        ANativeWindow_release(window->native_window);
+
+    av_free(window);
 
     return 0;
 }
diff --git a/libavcodec/mediacodec_surface.h b/libavcodec/mediacodec_surface.h
index 933dc2bf51..e2ac1c9057 100644
--- a/libavcodec/mediacodec_surface.h
+++ b/libavcodec/mediacodec_surface.h
@@ -25,10 +25,12 @@
 
 #include "libavcodec/avcodec.h"
 
-struct FFANativeWindow;
-typedef struct FFANativeWindow FFANativeWindow;
+typedef struct FFANativeWindow {
+    void *surface;
+    void *native_window;
+} FFANativeWindow;
 
-FFANativeWindow *ff_mediacodec_surface_ref(void *surface, void *log_ctx);
+FFANativeWindow *ff_mediacodec_surface_ref(void *surface, void *native_window, void *log_ctx);
 int ff_mediacodec_surface_unref(FFANativeWindow *window, void *log_ctx);
 
 #endif /* AVCODEC_MEDIACODEC_SURFACE_H */
diff --git a/libavcodec/mediacodec_wrapper.c b/libavcodec/mediacodec_wrapper.c
index b12aced711..284d615980 100644
--- a/libavcodec/mediacodec_wrapper.c
+++ b/libavcodec/mediacodec_wrapper.c
@@ -1371,12 +1371,17 @@ fail:
     return ret;
 }
 
-static int mediacodec_jni_configure(FFAMediaCodec* ctx, const FFAMediaFormat* format_ctx, void* surface, void *crypto, uint32_t flags)
+static int mediacodec_jni_configure(FFAMediaCodec *ctx,
+                                    const FFAMediaFormat* format_ctx,
+                                    FFANativeWindow* window,
+                                    void *crypto,
+                                    uint32_t flags)
 {
     int ret = 0;
     JNIEnv *env = NULL;
     FFAMediaCodecJni *codec = (FFAMediaCodecJni *)ctx;
     const FFAMediaFormatJni *format = (FFAMediaFormatJni *)format_ctx;
+    jobject *surface = window ? window->surface : NULL;
 
     JNI_GET_ENV_OR_RETURN(env, codec, AVERROR_EXTERNAL);
 
@@ -2151,16 +2156,27 @@ static int mediacodec_ndk_delete(FFAMediaCodec* ctx)
     return ret;
 }
 
-static int mediacodec_ndk_configure(FFAMediaCodec* ctx, const FFAMediaFormat* format_ctx, void* surface, void *crypto, uint32_t flags)
+static int mediacodec_ndk_configure(FFAMediaCodec* ctx,
+                                    const FFAMediaFormat* format_ctx,
+                                    FFANativeWindow* window,
+                                    void *crypto,
+                                    uint32_t flags)
 {
     FFAMediaCodecNdk *codec = (FFAMediaCodecNdk *)ctx;
     FFAMediaFormatNdk *format = (FFAMediaFormatNdk *)format_ctx;
     media_status_t status;
+    ANativeWindow *native_window = NULL;
 
-    if (surface) {
-        JNIEnv *env = NULL;
-        JNI_GET_ENV_OR_RETURN(env, ctx, -1);
-        codec->window = ANativeWindow_fromSurface(env, surface);
+    if (window) {
+        if (window->surface) {
+            JNIEnv *env = NULL;
+            JNI_GET_ENV_OR_RETURN(env, ctx, -1);
+            native_window = ANativeWindow_fromSurface(env, window->surface);
+            // Save for release
+            codec->window = native_window;
+        } else if (window->native_window) {
+            native_window = window->native_window;
+        }
     }
 
     if (format_ctx->class != &amediaformat_ndk_class) {
@@ -2168,7 +2184,7 @@ static int mediacodec_ndk_configure(FFAMediaCodec* ctx, const FFAMediaFormat* fo
         return AVERROR(EINVAL);
     }
 
-    status = codec->configure(codec->impl, format->impl, codec->window, NULL, flags);
+    status = codec->configure(codec->impl, format->impl, native_window, NULL, flags);
     if (status != AMEDIA_OK) {
         av_log(codec, AV_LOG_ERROR, "configure failed, %d\n", status);
         return AVERROR_EXTERNAL;
diff --git a/libavcodec/mediacodec_wrapper.h b/libavcodec/mediacodec_wrapper.h
index b9b882f243..7cf3f4aecd 100644
--- a/libavcodec/mediacodec_wrapper.h
+++ b/libavcodec/mediacodec_wrapper.h
@@ -27,6 +27,7 @@
 #include <sys/types.h>
 
 #include "avcodec.h"
+#include "mediacodec_surface.h"
 
 /**
  * The following API around MediaCodec and MediaFormat is based on the
@@ -163,7 +164,7 @@ struct FFAMediaCodec {
     FFAMediaCodec* (*createEncoderByType)(const char *mime_type);
     int (*delete)(FFAMediaCodec* codec);
 
-    int (*configure)(FFAMediaCodec* codec, const FFAMediaFormat* format, void* surface, void *crypto, uint32_t flags);
+    int (*configure)(FFAMediaCodec* codec, const FFAMediaFormat* format, FFANativeWindow* surface, void *crypto, uint32_t flags);
     int (*start)(FFAMediaCodec* codec);
     int (*stop)(FFAMediaCodec* codec);
     int (*flush)(FFAMediaCodec* codec);
@@ -202,7 +203,10 @@ FFAMediaCodec* ff_AMediaCodec_createCodecByName(const char *name, int ndk);
 FFAMediaCodec* ff_AMediaCodec_createDecoderByType(const char *mime_type, int ndk);
 FFAMediaCodec* ff_AMediaCodec_createEncoderByType(const char *mime_type, int ndk);
 
-static inline int ff_AMediaCodec_configure(FFAMediaCodec* codec, const FFAMediaFormat* format, void* surface, void *crypto, uint32_t flags)
+static inline int ff_AMediaCodec_configure(FFAMediaCodec *codec,
+                                           const FFAMediaFormat *format,
+                                           FFANativeWindow *surface,
+                                           void *crypto, uint32_t flags)
 {
     return codec->configure(codec, format, surface, crypto, flags);
 }
diff --git a/libavcodec/mediacodecdec_common.c b/libavcodec/mediacodecdec_common.c
index ae82efaa0a..80089439ea 100644
--- a/libavcodec/mediacodecdec_common.c
+++ b/libavcodec/mediacodecdec_common.c
@@ -589,14 +589,14 @@ int ff_mediacodec_dec_init(AVCodecContext *avctx, MediaCodecDecContext *s,
             if (device_ctx->type == AV_HWDEVICE_TYPE_MEDIACODEC) {
                 if (device_ctx->hwctx) {
                     AVMediaCodecDeviceContext *mediacodec_ctx = (AVMediaCodecDeviceContext *)device_ctx->hwctx;
-                    s->surface = ff_mediacodec_surface_ref(mediacodec_ctx->surface, avctx);
+                    s->surface = ff_mediacodec_surface_ref(mediacodec_ctx->surface, mediacodec_ctx->native_window, avctx);
                     av_log(avctx, AV_LOG_INFO, "Using surface %p\n", s->surface);
                 }
             }
         }
 
         if (!s->surface && user_ctx && user_ctx->surface) {
-            s->surface = ff_mediacodec_surface_ref(user_ctx->surface, avctx);
+            s->surface = ff_mediacodec_surface_ref(user_ctx->surface, NULL, avctx);
             av_log(avctx, AV_LOG_INFO, "Using surface %p\n", s->surface);
         }
     }