diff --git a/src/audio_core/adsp/apps/opus/opus_decoder.cpp b/src/audio_core/adsp/apps/opus/opus_decoder.cpp
index 35a9b3b828..2fd8013e35 100644
--- a/src/audio_core/adsp/apps/opus/opus_decoder.cpp
+++ b/src/audio_core/adsp/apps/opus/opus_decoder.cpp
@@ -93,7 +93,7 @@ public:
                 avc->request_sample_fmt = AV_SAMPLE_FMT_S16;
                 av_channel_layout_default(&avc->ch_layout, channel_count);
                 if (avcodec_open2(avc, codec, nullptr) >= 0) {
-                    avpkt = av_packet_alloc();
+                    pkt = av_packet_alloc();
                     frame = av_frame_alloc();
                     return ResultSuccess;
                 }
@@ -105,7 +105,7 @@ public:
 
     Result Shutdown() {
         av_frame_free(&frame);
-        av_packet_free(&avpkt);
+        av_packet_free(&pkt);
         avcodec_free_context(&avc);
         return ResultSuccess;
     }
@@ -125,17 +125,23 @@ public:
             while (rem_output_bytes > 0) {
                 int r = avcodec_receive_frame(avc, frame);
                 if (r == AVERROR(EAGAIN)) {
-                    av_packet_unref(avpkt);
-                    av_new_packet(avpkt, int(input_data_size) + AV_INPUT_BUFFER_PADDING_SIZE);
-                    std::memcpy(avpkt->data, reinterpret_cast<const u8*>(input_data), input_data_size);
-                    r = avcodec_send_packet(avc, avpkt);
+                    av_frame_unref(frame);
+                    av_packet_unref(pkt);
+                    if ((r = av_new_packet(pkt, int(input_data_size))) >= 0) {
+                        std::memcpy(pkt->data, reinterpret_cast<const u8*>(input_data), input_data_size);
+                        r = avcodec_send_packet(avc, pkt);
                     ASSERT(r >= 0);
+                    } else {
+                        return Service::Audio::ResultLibOpusInvalidState;
+                    }
                 } else if (r == AVERROR_EOF) {
+                    av_frame_unref(frame);
                     break;
                 } else if (r >= 0) {
-                    auto const bsize = av_samples_get_buffer_size(frame->linesize, frame->ch_layout.nb_channels, frame->nb_samples, AV_SAMPLE_FMT_S16, 1);
+                    auto const input_bsize = av_samples_get_buffer_size(frame->linesize, frame->ch_layout.nb_channels, frame->nb_samples, AV_SAMPLE_FMT_S16, 1);
                     if (frame->format == AV_SAMPLE_FMT_S16) {
-                        std::memcpy(reinterpret_cast<s16*>(output_data) + (int(output_data_size) - rem_output_bytes), frame->data[0], size_t(bsize));
+                        ASSERT(size_t(int(output_data_size) - rem_output_bytes + input_bsize) <= size_t(output_data_size));
+                        std::memcpy(reinterpret_cast<s16*>(output_data) + (int(output_data_size) - rem_output_bytes), frame->data[0], size_t(input_bsize));
                     } else {
                         SwrContext *swr = nullptr;
                         if (swr_alloc_set_opts2(
@@ -150,20 +156,19 @@ public:
                             nullptr
                         ) >= 0) {
                             if (swr_init(swr) >= 0) {
-                                AVFrame *s16_frame = av_frame_alloc();
-                                s16_frame->format = AV_SAMPLE_FMT_S16;
-                                s16_frame->sample_rate = frame->sample_rate;
-                                av_channel_layout_copy(&s16_frame->ch_layout, &frame->ch_layout);
-                                s16_frame->nb_samples = frame->nb_samples;
-                                av_frame_get_buffer(s16_frame, 0);
-                                swr_convert(swr, s16_frame->data, s16_frame->nb_samples, (const uint8_t **)frame->data, frame->nb_samples);
-                                std::memcpy(reinterpret_cast<s16*>(output_data) + (int(output_data_size) - rem_output_bytes), s16_frame->data[0], size_t(bsize));
-                                swr_free(&swr);
+                                int out_samples = int(av_rescale_rnd(swr_get_delay(swr, frame->sample_rate) + frame->nb_samples, frame->sample_rate, frame->sample_rate, AV_ROUND_UP));
+                                uint8_t *output = nullptr;
+                                av_samples_alloc(&output, frame->linesize, frame->ch_layout.nb_channels, out_samples, AV_SAMPLE_FMT_S16, 0);
+                                out_samples = swr_convert(swr, &output, out_samples, (const uint8_t **)frame->data, frame->nb_samples);
+                                std::memcpy(reinterpret_cast<s16*>(output_data) + (int(output_data_size) - rem_output_bytes), output, size_t(input_bsize));
+                                av_freep(&output);
                             }
+                                swr_free(&swr);
                         }
                     }
                     out_sample_count += frame->nb_samples;
-                    rem_output_bytes -= bsize;
+                    rem_output_bytes -= input_bsize;
+                    av_frame_unref(frame);
                 } else {
                     LOG_ERROR(Audio_DSP, "{}", r);
                     break;
@@ -176,7 +181,7 @@ public:
     }
 
     AVCodecContext* avc = nullptr;
-    AVPacket* avpkt = nullptr;
+    AVPacket* pkt = nullptr;
     AVFrame* frame = nullptr;
 };
 } // namespace