diff --git a/libavcodec/wmavoice.c b/libavcodec/wmavoice.c
index 7ddc434d78..a82ee946a5 100644
--- a/libavcodec/wmavoice.c
+++ b/libavcodec/wmavoice.c
@@ -134,7 +134,6 @@ typedef struct {
      * @name Global values specified in the stream header / extradata or used all over.
      * @{
      */
-    AVFrame frame;
     GetBitContext gb;             ///< packet bitreader. During decoder init,
                                   ///< it contains the extradata from the
                                   ///< demuxer. During decoding, it contains
@@ -443,9 +442,6 @@ static av_cold int wmavoice_decode_init(AVCodecContext *ctx)
     ctx->channel_layout         = AV_CH_LAYOUT_MONO;
     ctx->sample_fmt             = AV_SAMPLE_FMT_FLT;
 
-    avcodec_get_frame_defaults(&s->frame);
-    ctx->coded_frame = &s->frame;
-
     return 0;
 }
 
@@ -1733,7 +1729,8 @@ static int check_bits_for_superframe(GetBitContext *orig_gb,
  * @return 0 on success, <0 on error or 1 if there was not enough data to
  *         fully parse the superframe
  */
-static int synth_superframe(AVCodecContext *ctx, int *got_frame_ptr)
+static int synth_superframe(AVCodecContext *ctx, AVFrame *frame,
+                            int *got_frame_ptr)
 {
     WMAVoiceContext *s = ctx->priv_data;
     GetBitContext *gb = &s->gb, s_gb;
@@ -1801,13 +1798,13 @@ static int synth_superframe(AVCodecContext *ctx, int *got_frame_ptr)
     }
 
     /* get output buffer */
-    s->frame.nb_samples = 480;
-    if ((res = ff_get_buffer(ctx, &s->frame)) < 0) {
+    frame->nb_samples = 480;
+    if ((res = ff_get_buffer(ctx, frame)) < 0) {
         av_log(ctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return res;
     }
-    s->frame.nb_samples = n_samples;
-    samples = (float *)s->frame.data[0];
+    frame->nb_samples = n_samples;
+    samples = (float *)frame->data[0];
 
     /* Parse frames, optionally preceded by per-frame (independent) LSPs. */
     for (n = 0; n < 3; n++) {
@@ -1964,11 +1961,10 @@ static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
                 copy_bits(&s->pb, avpkt->data, size, gb, s->spillover_nbits);
                 flush_put_bits(&s->pb);
                 s->sframe_cache_size += s->spillover_nbits;
-                if ((res = synth_superframe(ctx, got_frame_ptr)) == 0 &&
+                if ((res = synth_superframe(ctx, data, got_frame_ptr)) == 0 &&
                     *got_frame_ptr) {
                     cnt += s->spillover_nbits;
                     s->skip_bits_next = cnt & 7;
-                    *(AVFrame *)data = s->frame;
                     return cnt >> 3;
                 } else
                     skip_bits_long (gb, s->spillover_nbits - cnt +
@@ -1983,12 +1979,11 @@ static int wmavoice_decode_packet(AVCodecContext *ctx, void *data,
     s->sframe_cache_size = 0;
     s->skip_bits_next = 0;
     pos = get_bits_left(gb);
-    if ((res = synth_superframe(ctx, got_frame_ptr)) < 0) {
+    if ((res = synth_superframe(ctx, data, got_frame_ptr)) < 0) {
         return res;
     } else if (*got_frame_ptr) {
         int cnt = get_bits_count(gb);
         s->skip_bits_next = cnt & 7;
-        *(AVFrame *)data = s->frame;
         return cnt >> 3;
     } else if ((s->sframe_cache_size = pos) > 0) {
         /* rewind bit reader to start of last (incomplete) superframe... */
diff --git a/libavcodec/ws-snd1.c b/libavcodec/ws-snd1.c
index 1c347e85a2..24ebcebfe0 100644
--- a/libavcodec/ws-snd1.c
+++ b/libavcodec/ws-snd1.c
@@ -41,28 +41,19 @@ static const int8_t ws_adpcm_4bit[] = {
      0,  1,  2,  3,  4,  5,  6,  8
 };
 
-typedef struct WSSndContext {
-    AVFrame frame;
-} WSSndContext;
-
 static av_cold int ws_snd_decode_init(AVCodecContext *avctx)
 {
-    WSSndContext *s = avctx->priv_data;
-
     avctx->channels       = 1;
     avctx->channel_layout = AV_CH_LAYOUT_MONO;
     avctx->sample_fmt     = AV_SAMPLE_FMT_U8;
 
-    avcodec_get_frame_defaults(&s->frame);
-    avctx->coded_frame = &s->frame;
-
     return 0;
 }
 
 static int ws_snd_decode_frame(AVCodecContext *avctx, void *data,
                                int *got_frame_ptr, AVPacket *avpkt)
 {
-    WSSndContext *s = avctx->priv_data;
+    AVFrame *frame     = data;
     const uint8_t *buf = avpkt->data;
     int buf_size       = avpkt->size;
 
@@ -89,18 +80,17 @@ static int ws_snd_decode_frame(AVCodecContext *avctx, void *data,
     }
 
     /* get output buffer */
-    s->frame.nb_samples = out_size;
-    if ((ret = ff_get_buffer(avctx, &s->frame)) < 0) {
+    frame->nb_samples = out_size;
+    if ((ret = ff_get_buffer(avctx, frame)) < 0) {
         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return ret;
     }
-    samples     = s->frame.data[0];
+    samples     = frame->data[0];
     samples_end = samples + out_size;
 
     if (in_size == out_size) {
         memcpy(samples, buf, out_size);
-        *got_frame_ptr   = 1;
-        *(AVFrame *)data = s->frame;
+        *got_frame_ptr = 1;
         return buf_size;
     }
 
@@ -176,9 +166,8 @@ static int ws_snd_decode_frame(AVCodecContext *avctx, void *data,
         }
     }
 
-    s->frame.nb_samples = samples - s->frame.data[0];
-    *got_frame_ptr   = 1;
-    *(AVFrame *)data = s->frame;
+    frame->nb_samples = samples - frame->data[0];
+    *got_frame_ptr    = 1;
 
     return buf_size;
 }
@@ -187,7 +176,6 @@ AVCodec ff_ws_snd1_decoder = {
     .name           = "ws_snd1",
     .type           = AVMEDIA_TYPE_AUDIO,
     .id             = AV_CODEC_ID_WESTWOOD_SND1,
-    .priv_data_size = sizeof(WSSndContext),
     .init           = ws_snd_decode_init,
     .decode         = ws_snd_decode_frame,
     .capabilities   = CODEC_CAP_DR1,
diff --git a/libavresample/audio_mix.c b/libavresample/audio_mix.c
index c701bf154b..fc37eac6a3 100644
--- a/libavresample/audio_mix.c
+++ b/libavresample/audio_mix.c
@@ -370,9 +370,6 @@ AudioMix *ff_audio_mix_alloc(AVAudioResampleContext *avr)
             goto error;
         av_freep(&avr->mix_matrix);
     } else {
-        int i, j;
-        char in_layout_name[128];
-        char out_layout_name[128];
         double *matrix_dbl = av_mallocz(avr->out_channels * avr->in_channels *
                                         sizeof(*matrix_dbl));
         if (!matrix_dbl)
@@ -399,27 +396,6 @@ AudioMix *ff_audio_mix_alloc(AVAudioResampleContext *avr)
             goto error;
         }
 
-        av_get_channel_layout_string(in_layout_name, sizeof(in_layout_name),
-                                     avr->in_channels, avr->in_channel_layout);
-        av_get_channel_layout_string(out_layout_name, sizeof(out_layout_name),
-                                     avr->out_channels, avr->out_channel_layout);
-        av_log(avr, AV_LOG_DEBUG, "audio_mix: %s to %s\n",
-               in_layout_name, out_layout_name);
-        av_log(avr, AV_LOG_DEBUG, "matrix size: %d x %d\n",
-               am->in_matrix_channels, am->out_matrix_channels);
-        for (i = 0; i < avr->out_channels; i++) {
-            for (j = 0; j < avr->in_channels; j++) {
-                if (am->output_zero[i])
-                    av_log(avr, AV_LOG_DEBUG, "  (ZERO)");
-                else if (am->input_skip[j] || am->output_skip[i])
-                    av_log(avr, AV_LOG_DEBUG, "  (SKIP)");
-                else
-                    av_log(avr, AV_LOG_DEBUG, "  %0.3f ",
-                           matrix_dbl[i * avr->in_channels + j]);
-            }
-            av_log(avr, AV_LOG_DEBUG, "\n");
-        }
-
         av_free(matrix_dbl);
     }
 
@@ -551,26 +527,13 @@ int ff_audio_mix_get_matrix(AudioMix *am, double *matrix, int stride)
     return 0;
 }
 
-int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
+static void reduce_matrix(AudioMix *am, const double *matrix, int stride)
 {
-    int i, o, i0, o0;
-
-    if ( am->in_channels <= 0 ||  am->in_channels > AVRESAMPLE_MAX_CHANNELS ||
-        am->out_channels <= 0 || am->out_channels > AVRESAMPLE_MAX_CHANNELS) {
-        av_log(am->avr, AV_LOG_ERROR, "Invalid channel counts\n");
-        return AVERROR(EINVAL);
-    }
-
-    if (am->matrix) {
-        av_free(am->matrix[0]);
-        am->matrix = NULL;
-    }
+    int i, o;
 
     memset(am->output_zero, 0, sizeof(am->output_zero));
     memset(am->input_skip,  0, sizeof(am->input_skip));
-    memset(am->output_skip, 0, sizeof(am->output_zero));
-    am->in_matrix_channels  = am->in_channels;
-    am->out_matrix_channels = am->out_channels;
+    memset(am->output_skip, 0, sizeof(am->output_skip));
 
     /* exclude output channels if they can be zeroed instead of mixed */
     for (o = 0; o < am->out_channels; o++) {
@@ -600,7 +563,7 @@ int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
     }
     if (am->out_matrix_channels == 0) {
         am->in_matrix_channels = 0;
-        return 0;
+        return;
     }
 
     /* skip input channels that contribute fully only to the corresponding
@@ -637,7 +600,7 @@ int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
     }
     if (am->in_matrix_channels == 0) {
         am->out_matrix_channels = 0;
-        return 0;
+        return;
     }
 
     /* skip output channels that only get full contribution from the
@@ -659,8 +622,31 @@ int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
     }
     if (am->out_matrix_channels == 0) {
         am->in_matrix_channels = 0;
-        return 0;
+        return;
     }
+}
+
+int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
+{
+    int i, o, i0, o0, ret;
+    char in_layout_name[128];
+    char out_layout_name[128];
+
+    if ( am->in_channels <= 0 ||  am->in_channels > AVRESAMPLE_MAX_CHANNELS ||
+        am->out_channels <= 0 || am->out_channels > AVRESAMPLE_MAX_CHANNELS) {
+        av_log(am->avr, AV_LOG_ERROR, "Invalid channel counts\n");
+        return AVERROR(EINVAL);
+    }
+
+    if (am->matrix) {
+        av_free(am->matrix[0]);
+        am->matrix = NULL;
+    }
+
+    am->in_matrix_channels  = am->in_channels;
+    am->out_matrix_channels = am->out_channels;
+
+    reduce_matrix(am, matrix, stride);
 
 #define CONVERT_MATRIX(type, expr)                                          \
     am->matrix_## type[0] = av_mallocz(am->out_matrix_channels *            \
@@ -686,6 +672,7 @@ int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
     }                                                                       \
     am->matrix = (void **)am->matrix_## type;
 
+    if (am->in_matrix_channels && am->out_matrix_channels) {
     switch (am->coeff_type) {
     case AV_MIX_COEFF_TYPE_Q8:
         CONVERT_MATRIX(q8, av_clip_int16(lrint(256.0 * v)))
@@ -700,6 +687,32 @@ int ff_audio_mix_set_matrix(AudioMix *am, const double *matrix, int stride)
         av_log(am->avr, AV_LOG_ERROR, "Invalid mix coeff type\n");
         return AVERROR(EINVAL);
     }
+    }
 
-    return mix_function_init(am);
+    ret = mix_function_init(am);
+    if (ret < 0)
+        return ret;
+
+    av_get_channel_layout_string(in_layout_name, sizeof(in_layout_name),
+                                 am->in_channels, am->in_layout);
+    av_get_channel_layout_string(out_layout_name, sizeof(out_layout_name),
+                                 am->out_channels, am->out_layout);
+    av_log(am->avr, AV_LOG_DEBUG, "audio_mix: %s to %s\n",
+           in_layout_name, out_layout_name);
+    av_log(am->avr, AV_LOG_DEBUG, "matrix size: %d x %d\n",
+           am->in_matrix_channels, am->out_matrix_channels);
+    for (o = 0; o < am->out_channels; o++) {
+        for (i = 0; i < am->in_channels; i++) {
+            if (am->output_zero[o])
+                av_log(am->avr, AV_LOG_DEBUG, "  (ZERO)");
+            else if (am->input_skip[i] || am->output_skip[o])
+                av_log(am->avr, AV_LOG_DEBUG, "  (SKIP)");
+            else
+                av_log(am->avr, AV_LOG_DEBUG, "  %0.3f ",
+                       matrix[o * am->in_channels + i]);
+        }
+        av_log(am->avr, AV_LOG_DEBUG, "\n");
+    }
+
+    return 0;
 }