diff --git a/modules/theora/video_stream_theora.cpp b/modules/theora/video_stream_theora.cpp
index 49346e36d9e..482fe001f93 100644
--- a/modules/theora/video_stream_theora.cpp
+++ b/modules/theora/video_stream_theora.cpp
@@ -504,28 +504,32 @@ void VideoStreamPlaybackTheora::update(double p_delta) {
 		ogg_packet op;
 
 		while (!audio_ready && !audio_done) {
+			// Send remaining frames
+			if (!send_audio()) {
+				audio_ready = true;
+				break;
+			}
+
 			float **pcm;
 			int ret = vorbis_synthesis_pcmout(&vd, &pcm);
 			if (ret > 0) {
-				const int AUXBUF_LEN = 4096;
-				int to_read = ret;
-				float aux_buffer[AUXBUF_LEN];
-				while (to_read) {
-					int m = MIN(AUXBUF_LEN / vi.channels, to_read);
+				int frames_read = 0;
+				while (frames_read < ret) {
+					int m = MIN(AUXBUF_LEN / vi.channels, ret - frames_read);
 					int count = 0;
 					for (int j = 0; j < m; j++) {
 						for (int i = 0; i < vi.channels; i++) {
-							aux_buffer[count++] = pcm[i][j];
+							audio_buffer[count++] = pcm[i][frames_read + j];
 						}
 					}
-					int mixed = mix_callback(mix_udata, aux_buffer, m);
-					to_read -= mixed;
-					if (mixed != m) { //could mix no more
+					frames_read += m;
+					audio_ptr_end = m;
+					if (!send_audio()) {
 						audio_ready = true;
 						break;
 					}
 				}
-				vorbis_synthesis_read(&vd, ret - to_read);
+				vorbis_synthesis_read(&vd, frames_read);
 			} else {
 				/* no pending audio; is there a pending packet to decode? */
 				if (ogg_stream_packetout(&vo, &op) > 0) {
@@ -649,6 +653,8 @@ void VideoStreamPlaybackTheora::seek(double p_time) {
 	audio_done = !has_audio;
 	theora_eos = false;
 	vorbis_eos = false;
+	audio_ptr_start = 0;
+	audio_ptr_end = 0;
 
 	ogg_stream_reset(&to);
 	if (has_audio) {
diff --git a/modules/theora/video_stream_theora.h b/modules/theora/video_stream_theora.h
index 89257ff59b4..97893c27a97 100644
--- a/modules/theora/video_stream_theora.h
+++ b/modules/theora/video_stream_theora.h
@@ -52,6 +52,11 @@ class VideoStreamPlaybackTheora : public VideoStreamPlayback {
 	Point2i size;
 	Rect2i region;
 
+	const static int AUXBUF_LEN = 4096;
+	float audio_buffer[AUXBUF_LEN];
+	int audio_ptr_start;
+	int audio_ptr_end;
+
 	int buffer_data();
 	int queue_page(ogg_page *page);
 	int read_page(ogg_page *page);
@@ -108,6 +113,20 @@ class VideoStreamPlaybackTheora : public VideoStreamPlayback {
 protected:
 	void clear();
 
+	_FORCE_INLINE_ bool send_audio() {
+		if (audio_ptr_end > 0) {
+			int mixed = mix_callback(mix_udata, &audio_buffer[audio_ptr_start * vi.channels], audio_ptr_end - audio_ptr_start);
+			audio_ptr_start += mixed;
+			if (audio_ptr_start == audio_ptr_end) {
+				audio_ptr_start = 0;
+				audio_ptr_end = 0;
+			} else {
+				return false;
+			}
+		}
+		return true;
+	}
+
 public:
 	virtual void play() override;
 	virtual void stop() override;