diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index d919363c66..eeb867332c 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -2486,7 +2486,12 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
                 uint8_t *const ptr = dest_y + block_offset[i];
                 const int dir      = h->intra4x4_pred_mode_cache[scan8[i]];
                 if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) {
-                    h->hpc.pred8x8l_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
+                    if (h->x264_build != -1) {
+                        h->hpc.pred8x8l_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
+                    } else
+                        h->hpc.pred8x8l_filter_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift),
+                                                        (h-> topleft_samples_available << i) & 0x8000,
+                                                        (h->topright_samples_available << i) & 0x4000, linesize);
                 } else {
                     const int nnz = h->non_zero_count_cache[scan8[i + p * 16]];
                     h->hpc.pred8x8l[dir](ptr, (h->topleft_samples_available << i) & 0x8000,
diff --git a/libavcodec/h264pred.c b/libavcodec/h264pred.c
index 3fc3bc76e3..49a19da4d4 100644
--- a/libavcodec/h264pred.c
+++ b/libavcodec/h264pred.c
@@ -558,6 +558,8 @@ av_cold void ff_h264_pred_init(H264PredContext *h, int codec_id,
     h->pred4x4_add  [ HOR_PRED   ]= FUNCC(pred4x4_horizontal_add          , depth);\
     h->pred8x8l_add [VERT_PRED   ]= FUNCC(pred8x8l_vertical_add           , depth);\
     h->pred8x8l_add [ HOR_PRED   ]= FUNCC(pred8x8l_horizontal_add         , depth);\
+    h->pred8x8l_filter_add [VERT_PRED   ]= FUNCC(pred8x8l_vertical_filter_add           , depth);\
+    h->pred8x8l_filter_add [ HOR_PRED   ]= FUNCC(pred8x8l_horizontal_filter_add         , depth);\
     if (chroma_format_idc <= 1) {\
     h->pred8x8_add  [VERT_PRED8x8]= FUNCC(pred8x8_vertical_add            , depth);\
     h->pred8x8_add  [ HOR_PRED8x8]= FUNCC(pred8x8_horizontal_add          , depth);\
diff --git a/libavcodec/h264pred.h b/libavcodec/h264pred.h
index 00af03a568..6708292048 100644
--- a/libavcodec/h264pred.h
+++ b/libavcodec/h264pred.h
@@ -101,6 +101,8 @@ typedef struct H264PredContext {
                           int16_t *block /*align 16*/, ptrdiff_t stride);
     void(*pred8x8l_add[2])(uint8_t *pix /*align  8*/,
                            int16_t *block /*align 16*/, ptrdiff_t stride);
+    void(*pred8x8l_filter_add[2])(uint8_t *pix /*align  8*/,
+                           int16_t *block /*align 16*/, int topleft, int topright, ptrdiff_t stride);
     void(*pred8x8_add[3])(uint8_t *pix /*align  8*/,
                           const int *block_offset,
                           int16_t *block /*align 16*/, ptrdiff_t stride);
diff --git a/libavcodec/h264pred_template.c b/libavcodec/h264pred_template.c
index fd1515b36d..3d67dd40ae 100644
--- a/libavcodec/h264pred_template.c
+++ b/libavcodec/h264pred_template.c
@@ -1124,6 +1124,83 @@ static void FUNCC(pred8x8l_horizontal_up)(uint8_t *_src, int has_topleft,
     SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
     SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
 }
+
+static void FUNCC(pred8x8l_vertical_filter_add)(uint8_t *_src, int16_t *_block, int has_topleft,
+                                     int has_topright, ptrdiff_t stride)
+{
+    int i;
+    pixel *src = (pixel*)_src;
+    const dctcoef *block = (const dctcoef*)_block;
+    pixel pix[8];
+
+    stride >>= sizeof(pixel)-1;
+
+    PREDICT_8x8_LOAD_TOP;
+
+    pix[0] = t0;
+    pix[1] = t1;
+    pix[2] = t2;
+    pix[3] = t3;
+    pix[4] = t4;
+    pix[5] = t5;
+    pix[6] = t6;
+    pix[7] = t7;
+
+    for(i=0; i<8; i++){
+        pixel v = pix[i];
+        src[0*stride]= v += block[0];
+        src[1*stride]= v += block[8];
+        src[2*stride]= v += block[16];
+        src[3*stride]= v += block[24];
+        src[4*stride]= v += block[32];
+        src[5*stride]= v += block[40];
+        src[6*stride]= v += block[48];
+        src[7*stride]= v +  block[56];
+        src++;
+        block++;
+    }
+
+    memset(_block, 0, sizeof(dctcoef) * 64);
+}
+
+static void FUNCC(pred8x8l_horizontal_filter_add)(uint8_t *_src, int16_t *_block, int has_topleft,
+                               int has_topright, ptrdiff_t stride)
+{
+    int i;
+    pixel *src = (pixel*)_src;
+    const dctcoef *block = (const dctcoef*)_block;
+    pixel pix[8];
+
+    stride >>= sizeof(pixel)-1;
+
+    PREDICT_8x8_LOAD_LEFT;
+
+    pix[0] = l0;
+    pix[1] = l1;
+    pix[2] = l2;
+    pix[3] = l3;
+    pix[4] = l4;
+    pix[5] = l5;
+    pix[6] = l6;
+    pix[7] = l7;
+
+    for(i=0; i<8; i++){
+        pixel v = pix[i];
+        src[0]= v += block[0];
+        src[1]= v += block[1];
+        src[2]= v += block[2];
+        src[3]= v += block[3];
+        src[4]= v += block[4];
+        src[5]= v += block[5];
+        src[6]= v += block[6];
+        src[7]= v +  block[7];
+        src+= stride;
+        block+= 8;
+    }
+
+    memset(_block, 0, sizeof(dctcoef) * 64);
+}
+
 #undef PREDICT_8x8_LOAD_LEFT
 #undef PREDICT_8x8_LOAD_TOP
 #undef PREDICT_8x8_LOAD_TOPLEFT