rgvolume is not adjusting volume?

[Cy]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

I used rganalysis to get peak (whatever that is) gain and reference
level on my songs. Now I'd like to apply that adjustment via rgvolume,
but I can't get any detectable difference to occur whether I use it or
not.

In my code I am creating a GstMessage using the values that rganalysis
put out earlier (loaded from a database) and a null source, then
gst_bus_post to post the message. My gst_bus_add_watch callback
displays those tags correctly during playback, giving me the impression
that they have been properly dispatched. And I have a rgvolume element
both added and linked right before alsasink.

But the volume remains unchanged. Loud songs are still too loud. Quiet
songs are still too quiet. For the song I'm testing the gain is 2.13,
peak 0.512525 and reference level 89. It should sound distinguishably
different when adjusted. But I can't hear any amplification going on.

I realize you can't just do this:
gst-launch-0.10 filesrc location=somewhere.mp3 ! audioconvert ! audioresample ! rganalysis ! rgvolume ! alsasink

...because the analysis cannot produce a result until the end of
stream. For that reason I already performed the analysis just to a
fakesink, and stored those results in a database. Supposedly I can then
produce those results as tags during the actual playback and have
rgvolume adjust the volume so that artificially dampened songs play
just as well as artificially amplified ones.

I can't use the GstTagSetter interface, because all the ones
implementing that are for encoding songs and saving them to file not
playing them out loud.

here's the code where I add the tags:
...
  gdouble gain = g_ascii_strtod(PQgetvalue(result,0,1),&end);
  gdouble peak = g_ascii_strtod(PQgetvalue(result,0,2),&end);
  gdouble level =  g_ascii_strtod(PQgetvalue(result,0,3),&end);

  if(!getenv("noreplaygain")) {
    static GstTagList* list = NULL;
    if(list==NULL)
      list = gst_tag_list_new();
    gst_tag_list_add(list, GST_TAG_MERGE_REPLACE,
                     GST_TAG_TRACK_GAIN, gain,
                     GST_TAG_TRACK_PEAK, peak,
                     GST_TAG_REFERENCE_LEVEL, level,
                     NULL);
   
    GstMessage* message = gst_message_new_tag(NULL,list);
    gst_bus_post(bus,message);
    //gst_message_unref(message);
  }
...
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBCAAGBQJPAlt1AAoJEF1ecRpMzcFybaEH/34AIaP2idC26aUWy8UyOuyN
O8oJ6/73lQXSVhFyD56jCKGeAwbSI+1++WxXc5Ukw1uizFzTgUp6e8G5lWauh13E
sLhBxe4KD3pZaRVgEWpFM3mLDvW7e1ea7QW88R6ACd63nns0NKLSQClOZqIIt3b1
m1LcXbUNv3YSS7QMbzEqROqmBtB+gDp3jAxxb9ENquv57aDbOfkS9USqJ9k7LU/k
u7RSodK1nCpQJWhEitfbstoER/6z0r3mRg/Jat9IDPX16XyHuFJ+Ci3GjD9GUPy2
BIno3a3OjOPYhNHUKHGlCfhdlAQEdAYtnNsPsmVtOog+nAtzyyyqMKEgsjWitbA=
=X/S3
-----END PGP SIGNATURE-----
_______________________________________________
gstreamer-devel mailing list
[hidden email]
http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel

[Cy]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Well, looks like I'm on my own then. *sighs*

I looked at the source of the gstrgvolume.c and gstrganalysis.c to see
if I could figure out how to notify rgvolume that it does indeed have a
gain and peak to apply. Since the rgvolume code silently failed when it
found no tags for said gain peak and level I added in a line to log a
message when it found no tags, but otherwise left the source well
unmodified.

I'm not really sure what the problem is here. I tried using
gst_bus_post to post a message with the tag list. I tried
gst_pad_push_event on the specific pad that reads those events. Oddly,
rgvolume's "sink" pad reads the events. Wouldn't the "src" pad be the
one that responded to events to adjust the gain and peak of the signal
filtering through? I tried gst_element_found_tags_for_pad on the
pipeline for that pad, and gst_element_found_tags on the pipeline in
general. In all 4 methods, the rgvolume should have found both a track
gain and a track level tag, yet the statement

if (!has_track_gain && !has_track_peak && !has_album_gain && !has_album_peak) {

...always returns true. The only thing I can think at this point is
that something discards without processing any and all tags on the bus
when the pipeline's state switches to playing. I tried switching the
state before and after pushing the tags, again with no luck in both
situations. Is there some event that I must wait for before the
rgvolume "sink" pad can accept events to set the gain and peak it uses?

Sure would be nice if you could just set them as properties.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBCAAGBQJPBV/NAAoJEF1ecRpMzcFyyAEH/0L+y9q21HUp1aVq6L1A7jgt
c/YyLSZAZhU4uUkiRqIo4wdIBepyh+Mrv2QePJ+lN33MsEklqu71TEQT2Zd2uYrA
+7Wwe+5AMzSl3+0w8DkA0lAyMOxdh2ETqNu83dSv9q0HBs479cJxiT29iDX8MsXT
UCNFl2DqlI6McPiYcbGfbOCYc4/5Y5zXCrd8m0TabP9NE4Ljtf2hQ2sbEn83mZNB
hHTU3g3LZhY+OnW0bXa2cYz1GVHwfYBL2Y2itw+9205+7EZg6Y8fl1LJf+4pzpAW
H54+CU09lZqrqibdZQgK+nqKoY9P4BUGUH0mTi2C9/PmCzfwuNJZVfO9Ai9s2/A=
=7552
-----END PGP SIGNATURE-----
_______________________________________________
gstreamer-devel mailing list
[hidden email]
http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel

[Cy]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

OK think I got it. If you call gst_pad_send_event the moment that pad
has been activated, on rgvolume's "sink" pad, then rgvolume will adjust
the volume. And conveniently there is a
gst_pad_set_activatepush_function to do just that. It doesn't take a
"user data" parameter though. :/ Also I'm not sure if the default value
of the function is important, but it's easy enough to extract the
function pointer and call it inside my own override.

So...
1) use rganalysis and watch the bus for the peak/gain/level tags
2) save them somewhere in a database or something per track/album
3) when playing the song, handle the activate push function on rgvolume's "sink"
4) in that handler, supply the tags rganalysis put out via gst_pad_send_event.
5) You can actually hear your quiet songs again!

It still doesn't seem to be damping loud songs... -5.9 dB gain at 0.1
peak was almost unnoticable. Or is damping greater at peaks closer to
1? I don't actually have very many loud songs heh...

Well, if I can figure anything out I'm doing wrong I'll continue to
post. Thanks for nothing and I hope this helps!
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)

iQEcBAEBCAAGBQJPBYFqAAoJEF1ecRpMzcFyaucH/0p+hybuVWeUA34dR/890sQe
B7Jmnx0jIuPyz71CYx7++xfTO/tiT/F/GB0iON1effgZo3ZEDYC60uuavlwCThMR
VyL4XOR50+evelb3r39cXSA5J5zmMytlKl2DO3N+xPghD3NVRfNHg/uZt/Zj1U5g
D481SGdU73+DPsyDj3DSABRuBPTPZNO0WaYpusw3g9b0FD09FzgRm9PSIyzNK9Jq
8eM8ZNv3imT7LQFmDOLsl4CXtNJD+sMzzGkmX8OxUIQS3iB+4T9qfGMzuVk3C//V
jKSlnCNgCRNAd8ndMBrihEtYNYYq64IwgkNAYc57UduNLFnD0JRrxX2Baz0IpHs=
=CIdQ
-----END PGP SIGNATURE-----
_______________________________________________
gstreamer-devel mailing list
[hidden email]
http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel