diff --git a/libavformat/oggparsespeex.c b/libavformat/oggparsespeex.c
index 14d2b38131..395ae59377 100644
--- a/libavformat/oggparsespeex.c
+++ b/libavformat/oggparsespeex.c
@@ -32,18 +32,25 @@
 
 struct speex_params {
     int final_packet_duration;
+    int seq;
 };
 
 static int speex_header(AVFormatContext *s, int idx) {
     struct ogg *ogg = s->priv_data;
     struct ogg_stream *os = ogg->streams + idx;
+    struct speex_params *spxp = os->private;
     AVStream *st = s->streams[idx];
     uint8_t *p = os->buf + os->pstart;
 
-    if (os->seq > 1)
+    if (!spxp) {
+        spxp = av_mallocz(sizeof(*spxp));
+        os->private = spxp;
+    }
+
+    if (spxp->seq > 1)
         return 0;
 
-    if (os->seq == 0) {
+    if (spxp->seq == 0) {
         int frames_per_packet;
         st->codec->codec_type = CODEC_TYPE_AUDIO;
         st->codec->codec_id = CODEC_ID_SPEEX;
@@ -70,6 +77,7 @@ static int speex_header(AVFormatContext *s, int idx) {
     } else
         vorbis_comment(s, p, os->psize);
 
+    spxp->seq++;
     return 1;
 }
 
@@ -90,11 +98,6 @@ static int speex_packet(AVFormatContext *s, int idx)
     struct speex_params *spxp = os->private;
     int packet_size = s->streams[idx]->codec->frame_size;
 
-    if (!spxp) {
-        spxp = av_mallocz(sizeof(*spxp));
-        os->private = spxp;
-    }
-
     if (os->flags & OGG_FLAG_EOS && os->lastpts != AV_NOPTS_VALUE &&
         os->granule > 0) {
         /* first packet of final page. we have to calculate the final packet
diff --git a/libavformat/oggparsevorbis.c b/libavformat/oggparsevorbis.c
index 2e8ea72129..618e6c4737 100644
--- a/libavformat/oggparsevorbis.c
+++ b/libavformat/oggparsevorbis.c
@@ -166,23 +166,24 @@ vorbis_header (AVFormatContext * s, int idx)
     struct ogg_stream *os = ogg->streams + idx;
     AVStream *st = s->streams[idx];
     struct oggvorbis_private *priv;
+    int pkt_type = os->buf[os->pstart];
 
-    if (os->seq > 2)
+    if (!(pkt_type & 1))
         return 0;
 
-    if (os->seq == 0) {
+    if (!os->private) {
         os->private = av_mallocz(sizeof(struct oggvorbis_private));
         if (!os->private)
             return 0;
     }
 
-    if (os->psize < 1)
+    if (os->psize < 1 || pkt_type > 5)
         return -1;
 
     priv = os->private;
-    priv->len[os->seq] = os->psize;
-    priv->packet[os->seq] = av_mallocz(os->psize);
-    memcpy(priv->packet[os->seq], os->buf + os->pstart, os->psize);
+    priv->len[pkt_type >> 1] = os->psize;
+    priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
+    memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
     if (os->buf[os->pstart] == 1) {
         const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
         unsigned blocksize, bs0, bs1;
@@ -224,7 +225,7 @@ vorbis_header (AVFormatContext * s, int idx)
             fixup_vorbis_headers(s, priv, &st->codec->extradata);
     }
 
-    return os->seq < 3;
+    return 1;
 }
 
 const struct ogg_codec ff_vorbis_codec = {