shithub: flite

Info  •  Files  •  Log

Merge pull request #14 from earboxer/pitch

* SSML: Ability to pass pitch and range prosody attributes

--- a/src/cg/cst_cg.c

+++ b/src/cg/cst_cg.c

@@ -445,15 +445,15 @@

     /* Smooth F0 and mark unvoice frames as 0.0 */

     cst_item *mcep;

     int i;

-    float mean, stddev;

+    float base_mean, base_stddev;

     /* cg_smooth_F0_naive(param_track); */

     cg_F0_interpolate_spline(utt,param_track);

-    mean = get_param_float(utt->features,"int_f0_target_mean", cg_db->f0_mean);

-    mean *= get_param_float(utt->features,"f0_shift", 1.0);

-    stddev = 

+    base_mean = get_param_float(utt->features,"int_f0_target_mean", cg_db->f0_mean);

+    base_mean *= get_param_float(utt->features,"f0_shift", 1.0);

+    base_stddev =

         get_param_float(utt->features,"int_f0_target_stddev", cg_db->f0_stddev);

 #if 0

     FILE *ftt; int ii;

@@ -469,6 +469,25 @@

     {

         if (voiced_frame(mcep))

         {

+            float mean = base_mean;

+            float stddev = base_stddev;

+            float local_f0_mean =

+            ffeature_float(mcep,

+                "R:mcep_link.parent.R:segstate.parent.R:SylStructure.parent.parent.R:Token.parent.local_f0_mean"

+            );

+            if (local_f0_mean != 0.0)

+            {

+                mean = local_f0_mean;

+            }

+            float local_f0_range =

+            ffeature_float(mcep,

+                "R:mcep_link.parent.R:segstate.parent.R:SylStructure.parent.parent.R:Token.parent.local_f0_range"

+            );

+            if (local_f0_range > 0.0)

+            {

+                /* feature_float returns 0 by default, shifted to allow 0 to be passed. */

+                stddev = local_f0_range - 1.0;

+            }

             /* scale the F0 -- which normally wont change it at all */

             param_track->frames[i][0] = 

                 (((param_track->frames[i][0]-cg_db->f0_mean)/cg_db->f0_stddev) 

--- a/src/synth/cst_ssml.c

+++ b/src/synth/cst_ssml.c

@@ -93,14 +93,22 @@

     while (!cst_streq(">",name))

     {

         /* I want names and values to be const */

-        if (i == 0)

+        fnn = "_name0";

+        vnn = "_val0";

+        // Tags with more than one attribute need to have additional

+        // attributes defined here.

+        if (cst_streq("volume", name))

         {

-            fnn="_name0"; vnn="_val0";

+            fnn = "_name1"; vnn = "_val1";

         }

-        else

+        else if (cst_streq("pitch", name))

         {

-            fnn="_name1"; vnn="_val1";

+            fnn = "_name2"; vnn = "_val2";

         }

+        else if (cst_streq("range", name))

+        {

+            fnn = "_name3"; vnn = "_val3";

+        }

 	if (cst_streq(name,"/"))

 	    feat_set_string(a,"_type","startend");

 	else

@@ -200,20 +208,29 @@

             if (cst_streq("rate",get_param_string(attributes,"_name0","")))

                 feat_set_float(word_feats,"local_duration_stretch",

                                1.0/feat_float(attributes,"_val0"));

-            if (cst_streq("rate",get_param_string(attributes,"_name1","")))

-                feat_set_float(word_feats,"local_duration_stretch",

-                               1.0/feat_float(attributes,"_val1"));

-            if (cst_streq("volume",get_param_string(attributes,"_name0","")))

-                feat_set_float(word_feats,"local_gain",

-                               feat_float(attributes,"_val0")/100.0);

+            // volume is stored in _name1

             if (cst_streq("volume",get_param_string(attributes,"_name1","")))

                 feat_set_float(word_feats,"local_gain",

                                feat_float(attributes,"_val1")/100.0);

+            // pitch is stored in _name2

+            if (cst_streq("pitch", get_param_string(attributes, "_name2", "")))

+            {

+                feat_set_float(word_feats, "local_f0_mean", feat_float(attributes, "_val2"));

+            }

+            // range is stored in _name3

+            if (cst_streq("range", get_param_string(attributes, "_name3", "")))

+            {

+                feat_set_float(word_feats, "local_f0_range",

+                               // shift by + 1.0 to allow 0.0 to be passed.

+                               feat_float(attributes, "_val3") + 1.0);

+            }

         }

         else if (cst_streq("end",feat_string(attributes,"_type")))

         {

             feat_remove(word_feats,"local_duration_stretch");

             feat_remove(word_feats,"local_gain");

+            feat_remove(word_feats, "local_f0_mean");

+            feat_remove(word_feats, "local_f0_range");

         }

     }