diff --git a/fftools/ffmpeg.h b/fftools/ffmpeg.h
index 786f925bc6..c61a670103 100644
--- a/fftools/ffmpeg.h
+++ b/fftools/ffmpeg.h
@@ -265,6 +265,8 @@ typedef struct InputFilterOptions {
 
 enum OFilterFlags {
     OFILTER_FLAG_DISABLE_CONVERT    = (1 << 0),
+    // produce 24-bit audio
+    OFILTER_FLAG_AUDIO_24BIT        = (1 << 1),
 };
 
 typedef struct OutputFilterOptions {
@@ -283,6 +285,9 @@ typedef struct OutputFilterOptions {
      */
     AVRational          output_tb;
 
+    AVDictionary       *sws_opts;
+    AVDictionary       *swr_opts;
+
     // A combination of OFilterFlags.
     unsigned            flags;
 
@@ -574,8 +579,6 @@ typedef struct OutputStream {
     OutputFilter *filter;
 
     AVDictionary *encoder_opts;
-    AVDictionary *sws_dict;
-    AVDictionary *swr_opts;
     char *apad;
 
     char *attachment_filename;
diff --git a/fftools/ffmpeg_filter.c b/fftools/ffmpeg_filter.c
index d2fd26af7e..8aa4053716 100644
--- a/fftools/ffmpeg_filter.c
+++ b/fftools/ffmpeg_filter.c
@@ -210,6 +210,9 @@ typedef struct OutputFilterPriv {
 
     AVRational              sample_aspect_ratio;
 
+    AVDictionary           *sws_opts;
+    AVDictionary           *swr_opts;
+
     // those are only set if no format is specified and the encoder gives us multiple options
     // They point directly to the relevant lists of the encoder.
     const int              *formats;
@@ -813,6 +816,17 @@ int ofilter_bind_ost(OutputFilter *ofilter, OutputStream *ost,
     if (!ofp->name)
         return AVERROR(EINVAL);
 
+    ret = av_dict_copy(&ofp->sws_opts, opts->sws_opts, 0);
+    if (ret < 0)
+        return ret;
+
+    ret = av_dict_copy(&ofp->swr_opts, opts->swr_opts, 0);
+    if (ret < 0)
+        return ret;
+
+    if (opts->flags & OFILTER_FLAG_AUDIO_24BIT)
+        av_dict_set(&ofp->swr_opts, "output_sample_bits", "24", 0);
+
     if (fgp->is_simple) {
         // for simple filtergraph there is just one output,
         // so use only graph-level information for logging
@@ -945,6 +959,8 @@ void fg_free(FilterGraph **pfg)
         OutputFilterPriv *ofp = ofp_from_ofilter(ofilter);
 
         av_frame_free(&ofp->fps.last_frame);
+        av_dict_free(&ofp->sws_opts);
+        av_dict_free(&ofp->swr_opts);
 
         av_freep(&ofilter->linklabel);
         av_freep(&ofilter->name);
@@ -1358,7 +1374,7 @@ static int configure_output_video_filter(FilterGraph *fg, AVFilterGraph *graph,
         snprintf(args, sizeof(args), "%d:%d",
                  ofp->width, ofp->height);
 
-        while ((e = av_dict_iterate(ost->sws_dict, e))) {
+        while ((e = av_dict_iterate(ofp->sws_opts, e))) {
             av_strlcatf(args, sizeof(args), ":%s=%s", e->key, e->value);
         }
 
@@ -1725,6 +1741,7 @@ static int configure_filtergraph(FilterGraph *fg, FilterGraphThread *fgt)
         return AVERROR(ENOMEM);
 
     if (simple) {
+        OutputFilterPriv *ofp = ofp_from_ofilter(fg->outputs[0]);
         OutputStream *ost = fg->outputs[0]->ost;
 
         if (filter_nbthreads) {
@@ -1738,17 +1755,17 @@ static int configure_filtergraph(FilterGraph *fg, FilterGraphThread *fgt)
                 av_opt_set(fgt->graph, "threads", e->value, 0);
         }
 
-        if (av_dict_count(ost->sws_dict)) {
-            ret = av_dict_get_string(ost->sws_dict,
+        if (av_dict_count(ofp->sws_opts)) {
+            ret = av_dict_get_string(ofp->sws_opts,
                                      &fgt->graph->scale_sws_opts,
                                      '=', ':');
             if (ret < 0)
                 goto fail;
         }
 
-        if (av_dict_count(ost->swr_opts)) {
+        if (av_dict_count(ofp->swr_opts)) {
             char *args;
-            ret = av_dict_get_string(ost->swr_opts, &args, '=', ':');
+            ret = av_dict_get_string(ofp->swr_opts, &args, '=', ':');
             if (ret < 0)
                 goto fail;
             av_opt_set(fgt->graph, "aresample_swr_opts", args, 0);
diff --git a/fftools/ffmpeg_mux.c b/fftools/ffmpeg_mux.c
index 253c2e58d4..557f08b3a5 100644
--- a/fftools/ffmpeg_mux.c
+++ b/fftools/ffmpeg_mux.c
@@ -819,9 +819,6 @@ static void ost_free(OutputStream **post)
 
     av_freep(&ost->attachment_filename);
 
-    av_dict_free(&ost->sws_dict);
-    av_dict_free(&ost->swr_opts);
-
     if (ost->enc_ctx)
         av_freep(&ost->enc_ctx->stats_in);
     avcodec_free_context(&ost->enc_ctx);
diff --git a/fftools/ffmpeg_mux_init.c b/fftools/ffmpeg_mux_init.c
index 219be5f965..8f4b73f8a7 100644
--- a/fftools/ffmpeg_mux_init.c
+++ b/fftools/ffmpeg_mux_init.c
@@ -1356,12 +1356,6 @@ static int ost_add(Muxer *mux, const OptionsContext *o, enum AVMediaType type,
     if (oc->oformat->flags & AVFMT_GLOBALHEADER && ost->enc_ctx)
         ost->enc_ctx->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
 
-    av_dict_copy(&ost->sws_dict, o->g->sws_dict, 0);
-
-    av_dict_copy(&ost->swr_opts, o->g->swr_opts, 0);
-    if (ost->enc_ctx && av_get_exact_bits_per_sample(ost->enc_ctx->codec_id) == 24)
-        av_dict_set(&ost->swr_opts, "output_sample_bits", "24", 0);
-
     MATCH_PER_STREAM_OPT(copy_initial_nonkeyframes, i,
                          ms->copy_initial_nonkeyframes, oc, st);
 
@@ -1392,10 +1386,13 @@ static int ost_add(Muxer *mux, const OptionsContext *o, enum AVMediaType type,
             .vsync_method = vsync_method,
             .sample_rate = ost->enc_ctx->sample_rate,
             .ch_layout   = ost->enc_ctx->ch_layout,
+            .sws_opts    = o->g->sws_dict,
+            .swr_opts    = o->g->swr_opts,
             .output_tb = enc_tb,
             .ts_offset = mux->of.start_time == AV_NOPTS_VALUE ?
                          0 : mux->of.start_time,
-            .flags = OFILTER_FLAG_DISABLE_CONVERT * !!keep_pix_fmt,
+            .flags = OFILTER_FLAG_DISABLE_CONVERT * !!keep_pix_fmt |
+                     OFILTER_FLAG_AUDIO_24BIT * !!(av_get_exact_bits_per_sample(ost->enc_ctx->codec_id) == 24),
         };
 
         snprintf(name, sizeof(name), "#%d:%d", mux->of.index, ost->index);