diff --git a/CHANGELOG.md b/CHANGELOG.md
index 7e4b70780..97634cedf 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -97,6 +97,7 @@
     Bug #5739: Saving and loading the save a second or two before hitting the ground doesn't count fall damage
     Bug #5758: Paralyzed actors behavior is inconsistent with vanilla
     Bug #5762: Movement solver is insufficiently robust
+    Bug #5807: Video decoding crash on ARM
     Bug #5821: NPCs from mods getting removed if mod order was changed
     Bug #5835: OpenMW doesn't accept negative values for NPC's hello, alarm, fight, and flee
     Bug #5836: OpenMW dialogue/greeting/voice filter doesn't accept negative Ai values for NPC's hello, alarm, fight, and flee
diff --git a/apps/openmw/mwsound/ffmpeg_decoder.cpp b/apps/openmw/mwsound/ffmpeg_decoder.cpp
index 95ed9eeed..84ea30b3c 100644
--- a/apps/openmw/mwsound/ffmpeg_decoder.cpp
+++ b/apps/openmw/mwsound/ffmpeg_decoder.cpp
@@ -287,9 +287,9 @@ void FFmpeg_Decoder::close()
     mStream = nullptr;
 
     av_packet_unref(&mPacket);
-    av_freep(&mFrame);
-    swr_free(&mSwr);
     av_freep(&mDataBuf);
+    av_frame_free(&mFrame);
+    swr_free(&mSwr);
 
     if(mFormatCtx)
     {
@@ -302,11 +302,9 @@ void FFmpeg_Decoder::close()
             //
             if (mFormatCtx->pb->buffer != nullptr)
             {
-                av_free(mFormatCtx->pb->buffer);
-                mFormatCtx->pb->buffer = nullptr;
+                av_freep(&mFormatCtx->pb->buffer);
             }
-            av_free(mFormatCtx->pb);
-            mFormatCtx->pb = nullptr;
+            avio_context_free(&mFormatCtx->pb);
         }
         avformat_close_input(&mFormatCtx);
     }
diff --git a/extern/osg-ffmpeg-videoplayer/audiodecoder.cpp b/extern/osg-ffmpeg-videoplayer/audiodecoder.cpp
index 337382679..c32794d2a 100644
--- a/extern/osg-ffmpeg-videoplayer/audiodecoder.cpp
+++ b/extern/osg-ffmpeg-videoplayer/audiodecoder.cpp
@@ -91,7 +91,7 @@ MovieAudioDecoder::~MovieAudioDecoder()
     if(mAudioContext)
         avcodec_free_context(&mAudioContext);
 
-    av_freep(&mFrame);
+    av_frame_free(&mFrame);
     av_freep(&mDataBuf);
 }
 
@@ -222,7 +222,7 @@ int MovieAudioDecoder::audio_decode_frame(AVFrame *frame, int &sample_skip)
             return result;
         }
 
-        av_packet_unref(&mPacket);
+        av_packet_unref(pkt);
         mGetNextPacket = true;
 
         /* next packet */
diff --git a/extern/osg-ffmpeg-videoplayer/videostate.cpp b/extern/osg-ffmpeg-videoplayer/videostate.cpp
index 5858d985a..a3a1e4318 100644
--- a/extern/osg-ffmpeg-videoplayer/videostate.cpp
+++ b/extern/osg-ffmpeg-videoplayer/videostate.cpp
@@ -2,6 +2,7 @@
 
 #include <algorithm>
 #include <cassert>
+#include <cstddef>
 #include <iostream>
 #include <thread>
 #include <chrono>
@@ -49,7 +50,7 @@ VideoState::VideoState()
     , av_sync_type(AV_SYNC_DEFAULT)
     , audio_st(nullptr)
     , video_st(nullptr), frame_last_pts(0.0)
-    , video_clock(0.0), sws_context(nullptr), rgbaFrame(nullptr), pictq_size(0)
+    , video_clock(0.0), sws_context(nullptr), pictq_size(0)
     , pictq_rindex(0), pictq_windex(0)
     , mSeekRequested(false)
     , mSeekPos(0)
@@ -82,10 +83,11 @@ void PacketQueue::put(AVPacket *pkt)
     pkt1 = (AVPacketList*)av_malloc(sizeof(AVPacketList));
     if(!pkt1) throw std::bad_alloc();
 
-    if(pkt != &flush_pkt && !pkt->buf && av_packet_ref(&pkt1->pkt, pkt) < 0)
-        throw std::runtime_error("Failed to duplicate packet");
+    if(pkt == &flush_pkt)
+        pkt1->pkt = *pkt;
+    else
+        av_packet_move_ref(&pkt1->pkt, pkt);
 
-    pkt1->pkt = *pkt;
     pkt1->next = nullptr;
 
     this->mutex.lock ();
@@ -116,7 +118,8 @@ int PacketQueue::get(AVPacket *pkt, VideoState *is)
             this->nb_packets--;
             this->size -= pkt1->pkt.size;
 
-            *pkt = pkt1->pkt;
+            av_packet_unref(pkt);
+            av_packet_move_ref(pkt, &pkt1->pkt);
             av_free(pkt1);
 
             return 1;
@@ -155,6 +158,39 @@ void PacketQueue::clear()
     this->mutex.unlock ();
 }
 
+int VideoPicture::set_dimensions(int w, int h) {
+  if (this->rgbaFrame != nullptr && this->rgbaFrame->width == w &&
+      this->rgbaFrame->height == h) {
+    return 0;
+  }
+
+  std::unique_ptr<AVFrame, VideoPicture::AVFrameDeleter> frame{
+      av_frame_alloc()};
+  if (frame == nullptr) {
+    std::cerr << "av_frame_alloc failed" << std::endl;
+    return -1;
+  }
+
+  constexpr AVPixelFormat kPixFmt = AV_PIX_FMT_RGBA;
+  frame->format = kPixFmt;
+  frame->width = w;
+  frame->height = h;
+  if (av_image_alloc(frame->data, frame->linesize, frame->width, frame->height,
+                     kPixFmt, 1) < 0) {
+    std::cerr << "av_image_alloc failed" << std::endl;
+    return -1;
+  }
+
+  this->rgbaFrame = std::move(frame);
+  return 0;
+}
+
+void VideoPicture::AVFrameDeleter::operator()(AVFrame* frame) const
+{
+    av_freep(frame->data);
+    av_frame_free(&frame);
+}
+
 int VideoState::istream_read(void *user_data, uint8_t *buf, int buf_size)
 {
     try
@@ -220,7 +256,7 @@ void VideoState::video_display(VideoPicture *vp)
         osg::ref_ptr<osg::Image> image = new osg::Image;
 
         image->setImage(this->video_ctx->width, this->video_ctx->height,
-                        1, GL_RGBA, GL_RGBA, GL_UNSIGNED_BYTE, &vp->data[0], osg::Image::NO_DELETE);
+                        1, GL_RGBA, GL_RGBA, GL_UNSIGNED_BYTE, vp->rgbaFrame->data[0], osg::Image::NO_DELETE);
 
         mTexture->setImage(image);
     }
@@ -296,23 +332,27 @@ int VideoState::queue_picture(AVFrame *pFrame, double pts)
     // Convert the image into RGBA format
     // TODO: we could do this in a pixel shader instead, if the source format
     // matches a commonly used format (ie YUV420P)
-    if(this->sws_context == nullptr)
+    const int w = pFrame->width;
+    const int h = pFrame->height;
+    if(this->sws_context == nullptr || this->sws_context_w != w || this->sws_context_h != h)
     {
-        int w = this->video_ctx->width;
-        int h = this->video_ctx->height;
+        if (this->sws_context != nullptr)
+            sws_freeContext(this->sws_context);
         this->sws_context = sws_getContext(w, h, this->video_ctx->pix_fmt,
                                            w, h, AV_PIX_FMT_RGBA, SWS_BICUBIC,
                                            nullptr, nullptr, nullptr);
         if(this->sws_context == nullptr)
             throw std::runtime_error("Cannot initialize the conversion context!\n");
+        this->sws_context_w = w;
+        this->sws_context_h = h;
     }
 
     vp->pts = pts;
-    vp->data.resize(this->video_ctx->width * this->video_ctx->height * 4);
+    if (vp->set_dimensions(w, h) < 0)
+        return -1;
 
-    uint8_t *dst[4] = { &vp->data[0], nullptr, nullptr, nullptr };
     sws_scale(this->sws_context, pFrame->data, pFrame->linesize,
-              0, this->video_ctx->height, dst, this->rgbaFrame->linesize);
+              0, this->video_ctx->height, vp->rgbaFrame->data, vp->rgbaFrame->linesize);
 
     // now we inform our display thread that we have a pic ready
     this->pictq_windex = (this->pictq_windex+1) % VIDEO_PICTURE_ARRAY_SIZE;
@@ -360,13 +400,11 @@ public:
     {
         VideoState* self = mVideoState;
         AVPacket pkt1, *packet = &pkt1;
+        av_init_packet(packet);
         AVFrame *pFrame;
 
         pFrame = av_frame_alloc();
 
-        self->rgbaFrame = av_frame_alloc();
-        av_image_alloc(self->rgbaFrame->data, self->rgbaFrame->linesize, self->video_ctx->width, self->video_ctx->height, AV_PIX_FMT_RGBA, 1);
-
         while(self->videoq.get(packet, self) >= 0)
         {
             if(packet->data == flush_pkt.data)
@@ -407,10 +445,7 @@ public:
 
         av_packet_unref(packet);
 
-        av_free(pFrame);
-
-        av_freep(&self->rgbaFrame->data[0]);
-        av_free(self->rgbaFrame);
+        av_frame_free(&pFrame);
     }
 
 private:
@@ -438,6 +473,7 @@ public:
 
         AVFormatContext *pFormatCtx = self->format_ctx;
         AVPacket pkt1, *packet = &pkt1;
+        av_init_packet(packet);
 
         try
         {
@@ -673,16 +709,13 @@ void VideoState::init(std::shared_ptr<std::istream> inputstream, const std::stri
         {
           if (this->format_ctx->pb != nullptr)
           {
-              av_free(this->format_ctx->pb->buffer);
-              this->format_ctx->pb->buffer = nullptr;
-
-              av_free(this->format_ctx->pb);
-              this->format_ctx->pb = nullptr;
+              av_freep(&this->format_ctx->pb->buffer);
+              avio_context_free(&this->format_ctx->pb);
           }
         }
         // "Note that a user-supplied AVFormatContext will be freed on failure."
         this->format_ctx = nullptr;
-        av_free(ioCtx);
+        avio_context_free(&ioCtx);
         throw std::runtime_error("Failed to open video input");
     }
 
@@ -756,11 +789,8 @@ void VideoState::deinit()
         ///
         if (this->format_ctx->pb != nullptr)
         {
-            av_free(this->format_ctx->pb->buffer);
-            this->format_ctx->pb->buffer = nullptr;
-
-            av_free(this->format_ctx->pb);
-            this->format_ctx->pb = nullptr;
+            av_freep(&this->format_ctx->pb->buffer);
+            avio_context_free(&this->format_ctx->pb);
         }
         avformat_close_input(&this->format_ctx);
     }
@@ -771,6 +801,11 @@ void VideoState::deinit()
         mTexture->setImage(nullptr);
         mTexture = nullptr;
     }
+
+    // Dellocate RGBA frame queue.
+    for (std::size_t i = 0; i < VIDEO_PICTURE_ARRAY_SIZE; ++i)
+        this->pictq[i].rgbaFrame = nullptr;
+
 }
 
 double VideoState::get_external_clock()
diff --git a/extern/osg-ffmpeg-videoplayer/videostate.hpp b/extern/osg-ffmpeg-videoplayer/videostate.hpp
index 641fce04b..015656084 100644
--- a/extern/osg-ffmpeg-videoplayer/videostate.hpp
+++ b/extern/osg-ffmpeg-videoplayer/videostate.hpp
@@ -95,7 +95,16 @@ struct VideoPicture {
     VideoPicture() : pts(0.0)
     { }
 
-    std::vector<uint8_t> data;
+    struct AVFrameDeleter {
+        void operator()(AVFrame* frame) const;
+    };
+
+    // Sets frame dimensions.
+    // Must be called before writing to `rgbaFrame`.
+    // Return -1 on error.
+    int set_dimensions(int w, int h);
+
+    std::unique_ptr<AVFrame, AVFrameDeleter> rgbaFrame;
     double pts;
 };
 
@@ -159,8 +168,8 @@ struct VideoState {
     double      video_clock; ///<pts of last decoded frame / predicted pts of next decoded frame
     PacketQueue videoq;
     SwsContext*  sws_context;
+    int sws_context_w, sws_context_h;
     VideoPicture pictq[VIDEO_PICTURE_ARRAY_SIZE];
-    AVFrame*     rgbaFrame; // used as buffer for the frame converted from its native format to RGBA
     int          pictq_size, pictq_rindex, pictq_windex;
     std::mutex pictq_mutex;
     std::condition_variable pictq_cond;