Allow the native formats of a channel to influence the audio that is going to the...
[asterisk/asterisk.git] / res / res_speech.c
index b902f40..e848b06 100644 (file)
@@ -159,17 +159,22 @@ int ast_speech_change(struct ast_speech *speech, char *name, const char *value)
 }
 
 /*! \brief Create a new speech structure using the engine specified */
-struct ast_speech *ast_speech_new(char *engine_name, int format)
+struct ast_speech *ast_speech_new(char *engine_name, int formats)
 {
        struct ast_speech_engine *engine = NULL;
        struct ast_speech *new_speech = NULL;
+       int format = AST_FORMAT_SLINEAR;
 
        /* Try to find the speech recognition engine that was requested */
        if (!(engine = find_engine(engine_name)))
                return NULL;
 
-       /* Make sure the requested format fits */
-       if (!(engine->formats & format))
+       /* Before even allocating the memory below do some codec negotiation, we choose the best codec possible and fall back to signed linear if possible */
+       if ((format = (engine->formats & formats)))
+               format = ast_best_codec(format);
+       else if ((engine->formats & AST_FORMAT_SLINEAR))
+               format = AST_FORMAT_SLINEAR;
+       else
                return NULL;
 
        /* Allocate our own speech structure, and try to allocate a structure from the engine too */